Research on Digital Reform Strategy and Teaching Platform Development for Plant Landscape Courses in Colleges and Universities for Smart Education

As an industry that promotes the harmonious coexistence of human beings and nature, landscape gardening has played a major historical role in building China’s ecological civilization and creating urban and rural habitat. With the continuous expansion of urban scale, the ecological and environmental problems faced by the discipline of landscape architecture have become more and more complex [1–2]. In the context of the in-depth implementation of science and education, talent and innovation-driven development strategy, landscape architecture education plays an important role in the field of human habitat construction, how to cultivate high-level applied talents with solid theoretical foundations, strong sense of innovation, and strong practical ability is a key topic of discussion in the relevant colleges and universities at this stage [3–5]. “Plant Landscape” is the core course for landscape architecture majors, which is characterized by strong cross-discipline, high composite degree, large content span and strong practicality. It aims to take garden plants as the design elements, through systematic theoretical learning and practical operation, to cultivate students’ ability to comprehensively utilize the knowledge related to garden plants to carry out plant landscape planning and design, and to have the skills to construct a reasonable garden plant landscape system in the context of large environment and large-scale city, and to effectively express the design ideas and contents [6–8]. There are many problems in traditional teaching, such as focusing on theory rather than practice, lecturing rather than experience, analyzing positively rather than thinking negatively, etc., which can no longer meet the needs of today’s industry [9–10]. At present, we are in the era of information digitalization explosion, in this context, college students face the explosion of various information every day, how to do a good job in line with the needs of the current era of college students’ education reform has become another challenge for college teachers [11–12]. Based on the concept of intelligent education, the digital teaching reform practice of the course “Plant Landscape” aims to take the learners’ needs as the main body, take the learners’ experience as the core, explore the curriculum design, integrate the high-quality educational resources, build an advanced educational platform, and provide efficient educational services [13–14].

In this paper, an online learning platform based on adaptive learning path recommendation algorithm is constructed to realize the digital reform of plant landscape courses in colleges and universities. Firstly, the knowledge points and courses are collectively called learning objects, in which the knowledge point model of the plant landscape course includes four important parameters: the difficulty of the knowledge points, the importance of the knowledge points, the learner’s mastery of the knowledge points, and the learner’s benefit from learning the knowledge points. Then we will explore the implementation process of personalized learning path recommendations based on the ant colony algorithm, and complete the design and implementation of an online learning platform. Finally, a college major in plant landscape is selected, student data is collected for learning path planning, and teaching experiments and questionnaire surveys are conducted.

2

Overview

With the continuous development of urban landscape, landscaping has attracted much attention, and plant landscape design and its related disciplines have developed rapidly. Literature [15] puts forward three suggestions to solve the problems existing in the current landscape plant curriculum, namely, establishing the teaching objective based on the cultivation of “identification-cultivation-application” ability, developing the teaching content of “ecology, innovation and expansion”, and promoting the cross-integration of landscape plant curriculum and planning and design curriculum, so as to cultivate landscape professionals with practical and innovative ability for the society. Literature [16] points out the common problems existing in the traditional curriculum system of landscape architecture in colleges and universities, and discusses the reform measures of this system from formulation to implementation, which provides reference value for the future research on the curriculum reform of landscape architecture. Literature [17] introduces the American vocational training KAS system in the curriculum reform of landscape architecture, and realizes the teaching reform of landscape design courses by establishing a scientific cycle of internal evaluation feedback system of landscape teaching, and then solves the problems of insufficient practical ability, irrational setting of teaching content and teaching resources in the process of cultivating talents in landscape architecture. Literature [18] takes landscape plant pathology course as an example, and puts forward corresponding reform measures for the problems existing in its teaching, such as unreasonable knowledge structure of teachers, short class time, many examples of agricultural plant diseases, low motivation of students, etc. The actual implementation verifies the feasibility of the proposed measures, improves the students’ motivation and their ability to analyze and solve problems, and helps to improve the It helps to improve the teaching quality of landscape plant pathology.

In addition, the literature [19] points out the outstanding problems in the teaching process of the landscape gardening course, and puts forward suggestions for course optimization in terms of basic theories, case studies, and course design, so that students can become more familiar with the various attributes of plants in landscape gardening and make the planting design more reasonable and skillful. Literature [20] emphasizes the importance of the integration of information technology and course teaching, and takes the landscape design course of Hunan Agricultural University as an example to analyze its teaching practice of information technology reform in terms of teaching design, teaching methods, and syllabus. Literature [21] deepens the reform through four curriculum reform sub-topics of landscape gardening undergraduate course teaching program adjustment, teaching content reconstruction, teaching organization reorganization, etc., in order to build a composite threedimensional network of landscape gardening undergraduate teaching content organization form framework, and then to achieve the overall cultivation goal of landscape gardening professionals. Literature [22] proposes an intelligent educational framework for landscape architecture based on the network model of landscape architecture technology, and verifies the scientificity and feasibility of the educational framework through qualitative analysis, which has important reference value for the sustainable development of future education in the discipline of landscape architecture.

3

Digital Teaching Platform for Higher Education Plant Landscape Courses

3.1

Learning Object Modeling

To perform learning path planning, it is first necessary to model what the learner is going to learn. In this section, the modeling approach for knowledge points and courses will be presented, which will be used for subsequent learning path planning [23].

3.1.1

Difficulty of knowledge points

Calculate the percentage of learners who were able to correctly answer the question corresponding to the knowledge point out of all the learners who participated in the study of the knowledge point and gave answers to the question corresponding to the knowledge point. The larger the percentage, the easier the question is. The difficulty of a question is then defined as follows: 1 $b_{i} = \frac{N_{n i}}{N_{A i}}$

Included among these

b_i for topic i difficulty;

N_ri is the number of learners who gave the correct answer to question i;

N_Ai is the number of all learners who have done topic i.

3.1.2

Level of learner mastery

Assuming that for a knowledge point M, there are m exercises examining the knowledge point, i.e., M = {P₁, P₁…P_m}, then the learner’s mastery of the knowledge point is defined as in Equation (2): 2 $P_{M} = \frac{E_{s}}{E_{A}}$

P_M is the learner’s mastery of the knowledge point M, E_S is the number of questions for which the learner can give the correct answer, and E_A is the total number of questions included in the knowledge point M . This calculation is simple and easy to understand, and is the method traditionally used in education - a student’s score is given by the number of questions answered correctly in a test paper, which measures the student’s mastery of the content at that stage of the learning process. This leads to the derivation of a method for calculating a learner’s gain from learning a particular point, as defined in equation (3): 3 $E_{p} = \frac{E_{e} - E_{s}}{E_{A} - E_{s}}$ included among these

E_e is the number of questions for which the learner was able to give the correct answer after studying the point;

E_s is the number of questions for which the learner will be able to give the correct answer when the point is about to be studied;

E_A is the total number of questions containing the knowledge point;

E_p is the benefit at the level of learning that the learner receives after learning the point.

3.1.3

Importance of knowledge points

The importance of a knowledge point is considered along three dimensions: the number of times the video corresponding to the knowledge point has been viewed repeatedly, the length of the instructional video corresponding to the knowledge point, and the frequency of the knowledge point appearing in the test questions. We consider that the importance of the knowledge point that has been viewed more times, the more test questions examining the knowledge point, and the longer the corresponding video is relatively higher one because the teacher devotes more space to teaching and examining the knowledge point, and the previous learners all put more effort on the corresponding knowledge point. Therefore knowledge point importance is defined as in equation (4): 4 $I m p_{M} = (\frac{Q_{M}}{Q_{A}} + \frac{L_{M}}{L_{A}}) (1 + \lg \frac{N_{s}}{N_{1}})$ included among these

Q_M is the number of questions examining knowledge point M;

Q_A is the total number of test questions included in the current course;

L_M is the total duration of the video containing KM M;

L_A is the total video duration of the current course;

N_s is the total number of learners who have watched the video containing the knowledge point at least once;

N_i is the total number of learners who have watched the video related to the knowledge point only once;

Imp_M is the level of importance of the knowledge point M.

3.1.4

Benefits to learners of learning the KMs

Two ways of calculating the gain of knowledge points are given: 1)

Based on the income calculation of learners’ learning ability and the difficulty of the questions, the same knowledge points have different effects for learners with different learning levels. According to the research results that have been completed by other members of the same group, the probability of learners answering the 0-1 scoring questions correctly is shown in equation (5): 5 $p_{i} (θ) = \int_{- \infty}^{(θ + 1 - b_{i})} \frac{e^{\frac{- z^{2}}{2}}}{\sqrt{2 π}} d z$

Included among these

p_i (θ) is the probability that a learner with a learning ability value of θ will answer question i correctly;

θ is the corresponding learning ability of the learner that.

b_i is the difficulty of topic i.

Calculating the contribution of answering a question correctly to a learner’s learning level is defined as in equation (6): 6 $I_{i} (θ) = \frac{{[p_{i}^{'} (θ)]}^{2}}{p_{i} (θ) (1 - p_{i} (θ))}$

Included among these

p_i (θ) is the probability that a learner with a learning ability value of θ will answer question i correctly;

p′_i(θ) is the first order derivative of p_i (θ).

Therefore the learning level improvement that a learner P with learning ability θ can get after learning a certain knowledge point M containing m test items is defined as in Equation (7): 7 $I_{M} (θ) = \sum_{i = 1}^{m} I_{i} (θ)$

The calculation of the learner’s learning ability θ is the result of the research that has been done by other members of the same group, which is calculated by first assuming that the answers given by the learner for each of the m questions of the knowledge point M are x₁, x₂, x₃ … x_m, and the probability that the student will get such a scoring pattern should be the product of the ability to give the correct answer for each question, i.e., $\prod_{i = 1}^{m} p (X_{i} = x_{i}; θ)$ , and the value of this product will change with the change of the θ. However the learner did get such a result, so there must exist an θ, such that $\prod_{i}^{m} p (X_{i} = x_{i}; θ)$ can be maximized, and by considering this probabilistic score as a function of θ, the function is defined as in Equation (8): 8 $L (θ) = L (x_{1}, x_{2}, x_{3} \dots x_{m}; θ) = \prod_{i = 1}^{m} p (X_{i} = x_{i}; θ)$

Obviously L(θ) is a likelihood function, then the calculation of the learning ability θ becomes a problem of finding the maximum value of the likelihood function L(θ).

2)

Adopting the previous learning behavior sequence-based exercise score prediction scheme, by analyzing the learner’s learning behavior sequence, it gives a learner’s answer to an exercise question, and then derives its mastery of a knowledge point. The calculation of a learner’s gain from learning a certain knowledge point is defined as in Equation (9): 9 $I_{M} (\bar{P}) = 1 - S_{M} (\bar{P})$ where $\bar{P}$ is the sequence of learning behaviors of the learner;

$I_{M} (\bar{P})$ is the gain that a learner with learning behavior sequence $\bar{P}$ gets from learning knowledge point M;

$S_{M} (\bar{P})$ is the ratio of the number of exercises under Knowledge Point M that a learner with Learning Behavior Sequence $\bar{P}$ can answer correctly to the total number of exercises under Knowledge Point M, calculated after Learning Behavior Sequence Analysis.

3.1.5

Learners’ mastery of the curriculum

The learner’s mastery of the course is quantified by using the learner’s mastery of all the knowledge points in the course. For a course T, the learner’s mastery of the course is defined as in Equation (10): 10 $P_{T} = \frac{1}{N} \sum_{i = 1}^{N} P_{M_{i}}$

Where P_τ is the learner’s mastery of course T, N is the number of knowledge points M contained in course T, and P_{M_i} is the learner’s mastery of knowledge point M in course T. According to the description of the calculation method of the mastery degree of knowledge points, it can be seen that P_{M_i} is a value between 0 and 1, when the learner has a complete mastery of the knowledge points, the value is 1; when the learner does not understand at all, the value is 0. Therefore, the total mastery degree of the N knowledge points will not be more than the N, and the range of the value of the P_T is also [0,1].

3.1.6

Importance of the curriculum

In traditional education, if a course is important, it is manifested by the fact that the vast majority of schools offering the relevant program make the course mandatory. The importance of a course can be quantified using the confidence level of the association rule, and the importance of a course is defined as in equation (11): 11 $I m p_{T} = C o n f (X_{T} = > Y_{T})$

Included among these

X_τ is a school with a major related to course T;

Y_T for a school that has course T and has course T as a required course for the major.

Conf (X_T => Y_T) is the confidence level of association rule X_T => Y_T.

3.1.7

Course learning benefits

The learning benefit of a course is the benefit that the learner can get after learning a course. The benefit that a learner can get from learning a knowledge point can be obtained by predicting the results of the learner’s answers to the exercises, and the course consists of a number of knowledge points, so the benefit that a learner can get from learning a course T can be defined as in Equation (12): 12 $I_{T} = \frac{1}{N} \sum_{i = 1}^{N} I_{M_{i}}$

Where N is the number of knowledge points contained in the course T, and I_{M_i} is the benefit that the learner can get after learning the knowledge point M_i contained in the course T.

3.2

Adaptive Learning Path Recommendations

Learning path recommendation based on plant landscape students’ user profiles is the core function of this system. The system connects to the domain knowledge ontology library, and generates the optimal learning path for plant landscape students based on the learner’s basic information, learner’s cognitive level, learner’s learning style, and based on the ant colony algorithm. The generation of learning paths is mainly divided into the following three parts: 1) Incoming learner’s personalized features: 2) Ontology inference rule setting: 3) Recommendation of learning paths based on ACO algorithm.

3.2.1

Incoming learner personalization characteristics

The data in the learner’s cognitive level and learning style submodels are important parameters for making recommendations regarding learning paths. In addition, the learner’s learning objectives are also necessary for making recommendations.

The cognitive level of the learner determines the degree of difficulty of the recommended resources, and the cognitive level parameter is the learner’s mastery of a certain knowledge point obtained when the learner takes the test, which is calculated by the method of fuzzy logic according to Bloom’s cognitive goal classification theory. If a learner logs into the system for the first time, the default cognitive level of the learner is the first level, which corresponds to the knowledge level in Bloom’s cognitive classification.

Learning style determines the type of learning resources suitable for the user. After entering the system, the user performs a learning style test to initialize his/her learning style, and then corrects his/her learning style based on the rule mining online learning behaviors, which ultimately results in the characteristics of the learner in four dimensions, forming a learning style parameter with a certain degree of dynamism.

In this study, the learner’s learning goal is a milestone for the knowledge points within this course, which is the starting point for reasoning, and its successor knowledge points are found according to the goal in order to make the recommendation of learning paths.

3.2.2

Ontological Reasoning Rule Setting

Since the relationship between the terms has been established in the domain ontology library, then the learner’s learning objective is passed in, and using this as a starting point to find the predecessor and related knowledge points of this knowledge point, the set of predecessor knowledge points of the learner’s learning objective can be obtained by using inference techniques, and then it is passed into the algorithmic processing module, which will further optimize the processing and ultimately generate the learning path [24]. The inference rules are as follows: 1)

(K1isbasedonk2)→(k2isbasisfork1) indicates that Knowledge Point 2 is the successor knowledge point of Knowledge Point 1, and in order to learn Knowledge Point 1, Knowledge Point 2 should be mastered beforehand.

2)

(K1ispartofk2)→(k2haspartk1) indicates that Knowledge Point 1 is part of Knowledge Point 2 and Knowledge Point 2 contains Knowledge Point 1.

3)

(K1relatetok2)→(k2relatetokl) indicates that Knowledge Point 1 is related to Knowledge Point 2 and can be learned in association as a content supplement.

The above rules describe the specific inference rules for the base, entailment, and correlation relations, respectively. Rule 1 describes the basis of the relationship, that is, to learn the knowledge point 1, we must first understand the knowledge point 2 before learning; Rule 2 describes the inclusion of the relationship, such as wanting to understand “why the rat in the zodiac in the first place”, we must first understand the legends about the zodiac; Rule 3 describes the correlation of the relationship between the two knowledge points have a certain degree of relevance. Rule 3 illustrates the correlation relationship, indicating that the two knowledge points have a certain degree of relevance and can be used as complementary content for learning.

3.2.3

Learning path recommendation based on ant colony algorithm

There are three main components involved in the ant colony algorithm: pheromone, heuristic information, and the probability of selecting the learning object. Heuristic information and pheromones are important parameters of this algorithm. The size of the value determines the final recommendation result. [25]. 1)

Heuristic information recognition

Learners with different knowledge levels and learning styles have different preferences for learning objects, so the heuristic information is determined by matching user characteristics and learning object attributes.

In this study, the similarity between vectors is calculated using Euclidean distance by matching learner characteristics and learning object attributes in two dimensions. During the calculation process, the smaller the value obtained as a result, the more similar the two are. Then the similarity calculation formula between the learner’s knowledge level and the learning object’s difficulty coefficient is as follows: 13 $d (k l, d l) = \sqrt{(\sum_{i = 1}^{n} | k l_{i} - d l_{i} |^{2})}$

Similarly, the similarity between the learner’s learning style and the learning object’s knowledge expression is calculated as follows: 14 $d (l s, r f) = \sqrt{(\sum_{i = 1}^{n} | l s_{i} - r f_{i} |^{2})}$

To summarize, the heuristic information in ACO recommendation algorithm can be expressed as: 15 $η_{i j} (t) = d (k l, d l) * d (l s, r f) = \sqrt{(\sum_{i = 1}^{n} | k l_{i} - d l_{i} |^{2})} * \sqrt{(\sum_{i = 1}^{n} | l s_{i} - r f_{i} |^{2})}$

2)

Pheromone Recognition

Following the biological principles of the ant colony algorithm, the pheromone in this study is expressed as the number of the number of users learning a particular learning object. In other words, the more users of the same category on a particular learning object on the learning path, then it means that this learning object is more suitable for that category of learners. The categories mentioned here are those that were classified in the learner categorization study in Chapter II.

In ACO algorithm, using m_ij(t) to denote the intensity of pheromone, $Δ m_{i j}^{a} (t)$ to denote the total number of learners, and $Δ m_{i j}^{s} (t)$ to denote the number of similar learners, the formula for pheromone updating is: 16 $m_{i j} (t + 1) = (1 - v) * m_{i j} (t) + Δ m_{i j}^{a} (t) + Δ m_{i j}^{s} (t)$ where v denotes the volatilization rate of the pheromone.

3)

Selection probability

According to the basic principle of the ant colony algorithm, corresponding to the basis of the inspirational information and pheromone set in this paper, the probability of the learner when choosing the next learning object is: 17 $p_{i j} (t) = \frac{n_{i j} (t) * m_{i j} (t)}{Σ η_{i j} (t) * m_{i j} (t)} = \frac{[\sqrt{(Σ_{i = 1}^{n} | k l_{i} - d l_{i} |^{2})} \sqrt{(Σ_{i = 1}^{n} | s_{i} - r f_{i} |^{2})}] * [(1 - ν) m_{i j} (t - 1) + Δ m_{i j}^{a} (t - 1) + Δ m_{i j}^{s} (t - 1)]}{Σ ([\sqrt{(Σ_{i = 1}^{n} | k l_{i} - d l_{i} |^{2})} \sqrt{(Σ_{i = 1}^{n} | l s_{i} - r f_{i} |^{2})}] * [(1 - ν) m_{i j} (t - 1) + Δ m_{i j}^{a} (t - 1) + Δ m_{i j}^{s} (t - 1)])}$

4)

Learning path recommendation algorithm description

The steps of the ant colony recommendation algorithm based on plant landscape student user profiles are as follows: Step 1:

Construct all possible learning paths from the library of Chinese learning domains of plant landscape middle school students according to the learning objectives of the current user;

Step2:

Initialize each parameter, which includes the degree of matching between the learner’s cognitive level and the difficulty level of the learning object d(kl, dl), the degree of matching between the learner’s style and the type of the learning object d(ls, rf ), and the degree of pheromone volatilization v;

Step3:

Obtain the cognitive level and learning style of the current learner r;

Step4:

Among the learners who have accomplished the same learning objectives, determine the neighboring users of the current learner based on the cognitive level and learning style;

Step5:

obtaining the learning path of each neighboring user based on the learning behavior records in the system;

Step6:

updating the pheromone based on the evaluation of neighboring users;

Step7:

calculating the selection probability of all possible next learning objects for the current learner based on the learning objects he/she has completed;

Step8:

Rank all possible path selection probabilities in descending order and select the path with the highest probability to recommend to the current learner.

3.3

Design and realization of online learning platform

3.3.1

Functional requirements analysis of the system

The online learning platform uses the internet as a platform, creating a new way of learning. Firstly, it fully utilizes modern computer technology, which creates a better working mechanism for the online learning platform; secondly, it creates a learning environment containing various types of learning resources for the learners, which substantially improves the convenience of the learners. The online learning platform developed in this paper can provide learners with a learning system that can be used anytime and anywhere, and this attribute lays the foundation for personalized online learning platform services and becomes an advantageous condition for this platform to attract learners to join. In addition, the e-learning platform developed in this paper can also provide learners with an efficient learning environment to maximize their knowledge within a certain time period. Three modules are designed in the online learning platform for personalized learning path recommendation, which are the learner module, the resource module, and the backend administrator module. Figure 1 shows the functional module diagram of the system. 1)

Learner module

The online learning platform constructed in this paper is based on a personalized learning path recommendation system and focuses on lifelong learners. The learner module includes login and registration, personal center, and personalized information collection.

Learners can firstly register and log in the system through their user names and passwords, and after entering the system, they can fill in or modify their basic information and account passwords in the personal center; secondly, learners can browse, collect and download their favorite course resources, view their recent learning records, upload their own characteristics of the target resources that need to be recommended by the system and fill in the learning style questionnaire.

2)

Resource Module

Resource module resource display, resource retrieval, resource details and personalized recommendation four functional sub-modules. Learners can either browse the learning resources they are interested in on their own or collect information in the personalized recommendation module to find the learning path they need. The resource details module includes resource introduction, collection, download, and other functions. The personalized recommendation function is the most important function in the e-learning platform, including the display of learning path diagram and learning path recommended resources. In this function, the learning path recommendation algorithm based on the two-dimensional feature model is used to collect the learner’s feature information and provide the learner with the learning path and its resources, which can be accessed to the course resource interface by directly clicking the link.

3)

Background administrator module

The main function of the background administrator is to carry out background management. To manage the registered user information and administrator information; to review the learning resources in the system, if the learning resources do not meet the requirements of the platform, appropriate increase, delete, change and check the operation; to query the comments in the system and delete the existence of some inappropriate remarks, and to repeatedly have inappropriate remarks on the user to warn or cancel the account of the operation.

3.3.2

System design objectives

The emergence of online learning platforms is to provide learners with a learning path without time and location constraints, saving learners’ time and being able to improve learners’ efficiency. The main goal of the online learning platform based on personalized learning path recommendation designed and developed in this study is to collect learners’ personalized information, including learning styles, learning goals, etc., and match it with the characteristics of the learning resources, so as to provide learners with a learning path and resources that meet their needs, and to reduce the “detour” of learners in the process of acquiring new knowledge. Therefore, the main functions of the online learning platform based on personalized learning path recommendation are: 1)

The system can provide learners with the knowledge map of the course, so that they can understand the course content conveniently and quickly.

2)

The Kolb Learning Style Measurement Scale has been established in the system, which can be used to investigate the learning styles of the learners; collect data on the learners’ learning tasks, learning hours, knowledge levels, etc., and synthesize the learning paths for the learners, so as to help them accomplish their learning goals more quickly and efficiently.

3)

The system can provide a catalog navigation bar so that learners can quickly locate the learning resources they need;

4)

The addition, deletion, modification and checking of learning resources in the system can be operated by the background administrator to ensure the accuracy of learning resources.

3.3.3

Overall system architecture design

In this paper, the system adopts the three-layer architecture pattern that is most used in current development, which are application layer, business logic layer and data access layer. Figure 2 shows the general architecture of the system. 1)

Application layer

The application layer is the uppermost structure of the interaction between the system and the user, which can realize the visualization of user input data, and the beautiful GUI interaction interface is a highlight of the application layer. In the e-learning platform, users can realize the functions of logging in and registering on the e-learning platform, displaying course resources, downloading, searching, etc. in the application layer.

2)

Business Logic Layer

The function of the business logic layer is to execute the logical operation of a specific problem and connect it to the data access layer after receiving the input information from the application layer. In the online learning platform, users are able to realize functions such as collecting user information and creating learning paths in the business logic layer.

3)

Data access layer

The data access layer is the main structure of the system interacting with the database, which is responsible for converting and storing various data formats and providing the basic services required by users. In the e-learning platform, the functions that users can realize in the data access layer include storing learner characteristic information, course resource information and so on.

4

Effectiveness of Digital Reform and Feedback of Higher Education Plant Landscape Courses

Student data were collected in a university plant landscape program, and the data were processed for cluster analysis and learning path planning. Two classes were selected for the teaching experiment. The experimental class used the learning path planning teaching platform constructed in this paper, while the control class was taught according to the original teaching tradition.

4.1

Potential knowledge state determination based on K-means algorithm

Before constructing the learning path, it is necessary to know the mastery of each attribute by each student. In the results of Study 2 we have given the probability of mastery of attributes for each individual based on the cognitive diagnostic model, followed by cluster analysis of the probability of mastery of attributes for all subjects in SPSS, cluster analysis is in fact an exploratory analytical method, which does not pre-set the criteria for classification in the process of performing cluster classification, but rather, it classifies the attributes on its own, starting from the sample data. In educational data mining, cluster analysis, as a basic data analysis method, is the most frequently used method in studying digital learning. Based on different classification criteria, cluster analysis can be divided into a variety of algorithms, including k-mean method, division method, hierarchical method, density algorithm, grid algorithm and so on. Among them, k-means algorithm is one of the most famous and widely used clustering algorithms, its main characteristics are easy to implement, concise, efficient to carry out, but also one of the most commonly used and typical algorithms, which adopts the distance as the evaluation index of similarity, that the closer the distance between two objects, the greater the similarity.

The advantage of k-means is that it can converge quickly and is easy to implement, and its core idea is to divide N data objects into M clusters, so that the sum of squares of data points within each class to the center of the class is minimized, the specific algorithm is as follows: 1)

Accept the input quantity n, then sample the data to randomly select n objects as initial clustering centers.

2)

For the remaining other objects, they are each assigned to the cluster to which they are most similar, based on their distance from the cluster center.

3)

The cluster center (the mean of all objects in that cluster) is then calculated for each of the resulting new clusters. This process is repeated until the standard measurement function begins to converge (the cluster centers no longer change).

In this study, the initial clustering center of 3 was chosen to start the exploratory cluster analysis of attribute mastery probability gradually, with the initial clustering value increasing, the magnitude of the change in the center of the clustering slowly slowed down, when the clustering value of 9, the center of the clustering does not exist to change, and convergence was achieved, and thus this study determines the final clustering center of 9 by the clustering analysis based on the k-means algorithm. The attribute mastery probability of each clustering center is shown in Table 1. Where A1 to B2 are the attributes. Attribute mastery probability is the externalization of students’ knowledge state, that is to say, attribute mastery probability expresses the potential knowledge state in specific continuous probability values. Through the mastery probability attribute, we can gain a clearer picture of students’ strengths and weaknesses in the problem-solving process.

Table 1.

Properties master the final cluster center of probability

Attribute	1	2	3	4	5	6	7	8	9
A1	0.938	0.921	0.918	0.923	0.922	0.927	0.922	0.922	0.922
A2	0.932	0.914	0.924	0.925	0.914	0.911	0.917	0.919	0.916
A3	0.935	0.922	0.916	0.917	0.916	0.923	0.914	0.925	0.921
A4	0.977	0.965	0.959	0.962	0.962	0.96	0.959	0.958	0.957
A5	0.978	0.958	0.963	0.962	0.968	0.961	0.964	0.969	0.964
A6	0.936	0.923	0.924	0.93	0.917	0.922	0.926	0.92	0.918
B1	0.952	0.938	0.936	0.938	0.933	0.934	0.941	0.928	0.94
B2	0.949	0.934	0.938	0.937	0.941	0.937	0.934	0.93	0.926

Considering that the attribute mastery probability and mastery pattern involved in this study are all based on the GDINA package analysis in the R software, therefore, it is still in accordance with 0.5 as the criterion for determining whether the attribute mastery probability in each of the above final clustering centers has mastered the attribute or not, which is greater than 0.5, i.e., it is the mastery of the attribute and is labeled as 1, while it is less than 0.5, i.e., it is the non-mastery of the attribute and is labeled as 0, which can be converted into Table 1 and converted into Table 2. Table 2 shows the attribute mastery patterns after clustering.

Table 2

The properties of the clustering master mode

	KS-1	KS-2	KS-3	KS-4	KS-5	KS-6	KS-7	KS-8	KS-9
A1	1	1	1	0	1	1	1	1	1
A2	1	1	1	1	1	0	1	0	1
A3	1	0	0	1	1	1	1	1	0
A4	0	1	1	1	0	1	1	1	1
A5	1	1	1	1	1	1	0	1	1
A6	0	0	0	1	1	1	0	1	0
B1	1	1	1	0	1	0	1	1	1
B2	1	0	1	1	1	1	1	1	1
Attribute	9	4	7	6	3	0	1	2	1
Number	143	64	92	69	43	36	51	54	91

Table 2 shows the attribute mastery patterns after clustering. From Table 2, we can see that, on the one hand, from the point of view of the number of attributes and the number of people in each attribute mastery pattern after clustering, there are 36 people in KS-6 who have not mastered all the knowledge attributes, which is the smallest number of people in all the categories, and the largest number of people is still in KS-1 who have mastered all the knowledge attributes, which has reached 143 people, and the second is in KS-3 who have mastered 6 knowledge attributes, which has 92 people; It is also found that the number of students followed by KS-9, which has only mastered 2 knowledge attributes, is 91. The number of students accounted for by the three clustering patterns of KS-4, KS-7, and KS-8 also varies and is not the same, with 69 students mastering the attribute pattern of KS-4, 51 students mastering the attribute pattern of KS-7, and 54 students mastering the attribute pattern of KS-8.

In conclusion, the above table shows that although the number of attributes examined in some clustered mastery patterns is the same, the specific knowledge attributes examined are not the same. In addition, the number of students who have mastered most of the examined knowledge attributes and the number of students who have not mastered most of the examined knowledge attributes do not differ much, which indicates that there is polarization in the mastery of the knowledge of this group of students, which is worthy of teachers to reflect on in the subsequent teaching work, and targeted teaching.

4.2

Planning of learning paths

The basic assumption of learning path construction is that students’ acquisition of knowledge or skills is gradual, i.e., the acquisition of knowledge is progressive. Then, this study will design learning paths based on the principle of inclusion relationship. In this case, the containment relationship is the final clustering center for comparing two patterns that have been transformed into attribute mastery patterns. Taking KS-4 and KS-5 in this study as an example, the mastery pattern vectors of KS-4, which have been clustered and transformed into mastery patterns, are (00010101) and the mastery pattern vector of KS-5 is (10010101). The two attribute mastery patterns are compared using the inclusion relationship:

From the above example, it can be seen that the mastery pattern of KS-4 is less than that of KS-5 (KS-4≦KS-5). Because KS-5 has mastered attribute A1 in addition to attributes A4, A6, B2, while KS-4 has not mastered attribute A1, but only A4, A6, B2, we say that the attribute mastery mode of KS-4 is smaller than that of KS-5, in other words, it is necessary to achieve the attribute mastery mode of KS-4 before it can be mastered in the attribute mastery mode of KS-5. In other words, the mastery pattern of KS-4 attribute must be achieved before the mastery pattern of KS-5 can be mastered.

Accordingly, a complete learning path can be obtained i.e. KS-6→ KS-4→ KS-5→ KS-3→ KS-1. After sorting according to the inclusion principle, the path diagram is obtained as shown in Fig. 3. In the figure, the lowest level is the attribute mastery pattern in which the number of mastery is at least 0, and the higher the level, the more mastery, and the highest level is the attribute mastery pattern in which the number of mastery is at most 8. In the figure, a indicates the number of attributes mastered in the attribute mastery pattern, b indicates the number of people mastering the attribute mastery pattern, and c indicates the content of the attributes mastered.

From Figure 3, it can be clearly seen that there are complete 4 learning paths from the case of not mastering all the examined knowledge attributes to the case of mastering all the examined knowledge attributes, in which the longest learning path includes 5 potential knowledge states after clustering, and the shortest learning path includes 3 potential knowledge states after clustering. Meanwhile, by counting the number of potential knowledge states in each path, the number of students in these 4 complete learning paths can be obtained, as shown in Table 3.

Table 3.

Complete learning path

Categories	Path	Overall population
1	KS-6→ KS-9→ KS-8→ KS-1	329
2	KS-6→ KS-9→ KS-7→ KS-1	314
3	KS-6→ KS-4→ KS-5→ KS-3→ KS-1	387
4	KS-6→ KS-2→ KS-1	243

4.3

Analysis of questionnaire results

Two classes are selected for teaching experiments. The experimental class adopts the learning path planning teaching platform built in this paper, while the control class teaches according to the original teaching tradition.

In this study, we collected data from the experimental and traditional control groups and analyzed them using SPSS.

4.3.1

Comparison of learning outcomes

The t-test was used to examine the differences in the dimensions of personalized learning objectives between the two different teaching modes, namely, blended teaching based on the Learning Path Planning Platform and traditional teaching. The results show that the teaching activities of the different modes show significant differences at the 0.01 level (p=0.00<0.01) in the dimensions of knowledge and skills, process and method, and affective attitudes of learners. (p=0.00<0.01), and by comparing the mean values of 3.90>3.57, 3.93>3.14, 3.83>3.21, it is found that the mean value of the classes taught by the Learning Path Planning Platform is larger than that of the traditional teaching classes, which can be seen that compared with the traditional teaching mode, the learning path planning platform is more effective than the traditional teaching mode, and the learning path planning platform is more effective than the traditional teaching mode. It can be seen that compared with the traditional teaching mode, the blended teaching mode based on the “learning path planning platform” is more capable of promoting students’ personalized learning. This is because the personalized teaching mode based on the Learning Path Planning Platform starts from the problem situation, guides students to carry out independent and cooperative inquiry learning, and pays attention to the independent development of students in the learning process, which can stimulate students’ positive emotional participation and better contribute to the achievement of teaching goals.

Table 4.

Analysis of the difference of personalized learning

Dimension	Class	Number	Mean	SD	T
Knowledge and skills	Experimental Class	150	3.9	0.776	0.047
Knowledge and skills	Control Class	150	3.57	0.6624	0.047
Process and method	Experimental Class	150	3.93	0.8116	0.039
Process and method	Control Class	150	3.14	0.6125	0.039
Emotional attitude	Experimental Class	150	3.83	0.6941	0.034
Emotional attitude	Control Class	150	3.21	0.6225	0.034

4.3.2

Relevance of learning needs to learning objectives

1)

Satisfaction of personalized content needs is conducive to personalized learning objectives

The learners’ content needs and personalized learning objectives are tested for correlation analysis, and the Pearson correlation coefficient is used to indicate the strength of the correlation. First of all, the P-value of each dimension of content needs and the overall personalized learning objectives, which is also the Sig value in the table, is less than 0.01, which proves that the learners’ content needs and personalized learning objectives show significance at the 0.01 level. And the R-values are all positive, indicating that they are positively correlated. The specific correlation coefficient values of each dimension of content needs and each dimension of personalized learning objectives are shown in Table 5, and the R values are all in the range of 0.6 to 0.9, indicating that there is a close correlation between learning activities and personalized learning objectives.

2)

The satisfaction of personalized resource needs is conducive to the realization of personalized learning objectives

The learners’ resource needs and personalized learning goals are tested in correlation analysis, and the results of Pearson correlation coefficient test are shown in Table 6. First of all, the P-value (Sig value) of the correlation between the dimensions of resource type, amount of resources and learning partners under the dimension of learners ’ resource needs and personalized learning goals is <0.05, which proves that the learners ’ resource needs and personalized learning goals show significance at the 0.05 level. And the R-values are all positive, indicating a significant positive correlation between the two. The correlation coefficients between the dimensions of resource type, resource amount and learning partners and the dimensions of personalized learning goals are 0.5<R value<0.7, indicating that the influence of resource needs, including resource type, resource amount and learning partners, on personalized learning goals is relatively strong.

3)

The fulfillment of personalized process needs is conducive to the achievement of personalized learning goals

The dimensions of learners’ process needs in learning are correlated with personalized learning goals, and the findings are shown in Table 7. The results show that there is a high correlation between the satisfaction of learners ’ process needs and the achievement of personalized learning goals. First of all, the p-value (Sig value) = 0.000<0.01 for the dimensions of learning mode, learning progress, interaction mode, and interaction frequency under the dimension of learner’s process needs and the dimensions of personalized learning goals, which proves that the process needs of the learners and the personalized learning goals show significance at the 0.01 level. And the correlation coefficients R values are all positive, indicating a significant positive correlation between the two. The correlation coefficients of the dimensions of learning style, learning progress, interaction style, and interaction frequency with the dimensions of personalized learning goals have an R value greater than 0.7, which indicates that there is a close influence between the process needs and personalized learning goals, and in particular a high correlation is shown for the learner’s affective attitudes.

4)

Satisfaction of personalized evaluation needs is conducive to the achievement of personalized learning goals

The correlation analysis between the dimensions of evaluation needs in the teaching of “Learning Path Planning Platform” and the dimensions of personalized learning objectives reveals that there is a correlation between the evaluation needs in the teaching and the achievement of personalized learning objectives, and the specific results are shown in the table below. First of all, the P-value of the dimensions of evaluation needs and personalized learning objectives is <0.01, which proves that the evaluation needs of learners and personalized learning objectives show significance at the 0.01 level. And the R-values are all positive, indicating that they are positively correlated. Its correlation coefficient 0.6<R value<0.8, indicating that there is a correlation between evaluation needs and personalized learning objectives. The R-value of appraisal mode and evaluation feedback is >0.7, which means that appraisal mode and evaluation feedback will have a significant effect on achieving personalized learning objectives.

Table 5.

Correlation analysis of content requirements and learning goals

Dimension		Knowledge and skills		Process and method		Emotional attitude
Dimension		Coefficient	Sig.(Double tail)	Coefficient	Sig.(Double tail)	Coefficient	Sig.(Double tail)
Content demand	Content difficulty	0.757**	0.000	0.689**	0.004	0.892**	0.000
	Content span	0.691**	0.003	0.625**	0.003	0.834**	0.000
	Learning activity	0.878**	0.000	0.706**	0.000	0.768**	0.000

Table 6.

Correlation analysis of resource requirements and learning goals

Dimension		Knowledge and skills		Process and method		Emotional attitude
Dimension		Coefficient	Sig.(Double tail)	Coefficient	Sig.(Double tail)	Coefficient	Sig.(Double tail)
Resource requirements	Resource type	0.552*	0.014	0.507*	0.037	0.689**	0.000
	Resource quantity	0.657**	0.000	0.645**	0.000	0.621**	0.000
	Learning partner	0.592**	0.000	0.536*	0.029	0.542*	0.029

Table 7.

The correlation analysis of process requirements and learning goals

Dimension		Knowledge and skills		Process and method		Emotional attitude
Dimension		Coefficient	Sig.(Double tail)	Coefficient	Sig.(Double tail)	Coefficient	Sig.(Double tail)
Process demand	Learning mode	0.804**	0	0.816**	0	0.744**	0
	Learning schedule	0.743**	0	0.775**	0	0.849**	0
	Interactive mode	0.825**	0	0.724**	0	0.805**	0
	Interaction frequency	0.798**	0	0.787**	0	0.781**	0

Table 8.

Evaluate the relevance of requirements to learning goals

Dimension		Knowledge and skills		Process and method		Emotional attitude
Dimension		Coefficient	Sig.(Double tail)	Coefficient	Sig.(Double tail)	Coefficient	Sig.(Double tail)
Evaluation requirement	Evaluation criteria	0.675**	0	0.618**	0.006	0.613**	0
	Assessment mode	0.788**	0	0.702**	0	0.719**	0
	Evaluation time	0.607**	0	0.664**	0	0.662**	0.004
	Evaluation feedback	0.873**	0	0.735**	0	0.742**	0

5

Conclusion

In this paper, an online learning platform based on an adaptive learning path recommendation algorithm is constructed to facilitate the digital reform of plant landscape courses in colleges and universities. Plant landscape majors at a university were selected, and student data was collected for learning path planning applications and teaching experiments. After clustering analysis, there are 36 KS-6 students who have not mastered all the knowledge attributes, and the largest number is KS-1 students who have mastered all the knowledge attributes, reaching 143 students. Accordingly, it can be known that even though some clustering mastery patterns examine the same number of attributes, the specific knowledge attributes examined are not the same, and it is worthwhile for teachers to reflect on the subsequent teaching work and carry out the teaching work in a targeted manner. The learning path is thus obtained: KS-6→ KS-4→ KS-5→ KS-3→ KS-1. The teaching activities of different modes show significant differences at the 0.01 level in the dimensions of learners’ “knowledge and skills”, “process and method” and “affective attitude” (p=0.00<0.01), and it is found that the comparison of the means of the “learning path planning platform” is 3.90>3.57, 3.93>3.14, 3.83>3.21. 0.00<0.01), and through the comparison of the mean values of 3.90>3.57, 3.93>3.14, 3.83>3.21, it is found that the mean values of the classes taught by the Learning Path Planning Platform are greater than those of the traditional teaching classes, which can be seen that compared with the traditional teaching mode, the learning path planning platform is based on the “Learning Path Planning Platform”. It can be seen that compared to the traditional teaching mode, the blended teaching mode based on the Learning Path Planning Platform is more capable of promoting students’ personalized learning. This is due to the fact that the personalized teaching mode of the Learning Path Planning Platform starts from problematic situations, guides students to carry out independent and cooperative inquiry learning, and focuses on the independent development of students in the learning process, which can stimulate students’ positive emotional participation and better contribute to the achievement of teaching goals.

Language:: English

Publication timeframe:: 1 times per year
Journal Subjects:: Life Sciences, Life Sciences, other, Mathematics, Applied Mathematics, General Mathematics, Physics, Physics, other

Journal RSS Feed

Research on Digital Reform Strategy and Teaching Platform Development for Plant Landscape Courses in Colleges and Universities for Smart Education

Hui Zhang

Published Online: Mar 19, 2025

Received: Oct 19, 2024

Accepted: Feb 08, 2025

DOI: https://doi.org/10.2478/amns-2025-0449

KeywordsPath planning, Ontological reasoning, Ant colony algorithm, Teaching reforms

© 2025 Hui Zhang, published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Keywords
Path planning, Ontological reasoning, Ant colony algorithm, Teaching reforms