An Optimization Study of Information Visualization Techniques in Educational Management Systems for Student Data Analysis

Missing value analysis. Firstly, analyze the reasons for missing data, such as the inability to obtain unstructured data or the process of obtaining structured data is wrong; secondly, analyze the impact of missing data, for example, data mining algorithm models will lose a lot of useful information, it is difficult to grasp the laws embedded within the data; finally, use the Pandas library of Python to read in the collected raw data to view the basic situation of the data and calculate the the number of missing, not missing, missing rate and so on.

Outlier Analysis. Outlier analysis, also known as outlier analysis, refers to the individual values that exist in the raw data that significantly deviate from the rest of the observations, and is mainly processed using statistical volume analysis and box plot analysis, and the box plot can detect outliers, as shown in Figure 1. 1)

Statistical quantitative analysis is to do a descriptive analysis of the data to determine the abnormal data, for example, by determining the maximum and minimum values of the data in the data set to determine whether the value of the data is outside a reasonable range.

Figure 1.

Box diagram detects the exception value

Data integration refers to the process of logically or physically centralizing the storage of data sources with different data contents, data formats and data quality, and its main purpose is to solve the problem of data distribution and heterogeneity. In the process of data integration, due to the different ways of storing data in multiple data sources and the mismatch of data types, the original data are converted, refined and integrated after giving full consideration to the problems of data object identification and attribute redundancy. 1)

Identification of data objects

Data object identification refers to the identification of existing data objects from data sources with different data contents and data formats, and the common forms of stored data include three aspects: i.e., homonym, synonym and unit inconsistency. Among them, homonymy means that the same behavioral data in different data sources are used to describe different data objects; heteronymy means that the same data object in different data sources is described by different behavioral data; and unit non-uniformity means that the behavioral data of the same data object is described by using the international unit and the traditional unit of measurement respectively. Therefore, the purpose of identifying data objects is to detect and solve the problem of storing data forms in different data sources.

In order to unify the different magnitudes that exist between the behavioral data of the data object and the range differences of the behavioral data of the data object, it is necessary to normalize the data object and scale the behavioral data of the data object so that the values are distributed in a specific region, which facilitates the comprehensive analysis of the data object. 1)

Normalization of deviation

Off-diagonal normalization refers to min-max normalization, which refers to mapping the behavioral data of the behavioral data object of the linear transformation processing data object in the interval [0,1]. The off-diagonal normalization retains the original relationship between the behavioral data of the behavioral data object, and is the simplest way to eliminate the influence of the magnitude and range of values of the behavioral data of the behavioral data object, and the conversion method is shown in Equation (1). 1x*=x−minmax−min

where x^* denotes the behavioral data value of the behavioral data object after standard deviation normalization; x denotes the initial value of the behavioral data; x¯ denotes the mean value of the behavioral data: σ denotes the standard deviation of the behavioral data.

The essence of continuous discretization of behavioral data is to determine the split points of behavioral data values and map the values of behavioral data to corresponding subintervals. Commonly used methods include the equal-width split-box method, equal-frequency split-box method, and methods based on cluster analysis. 1)

Equal-width binning method

The equal-width split-box method divides continuous behavioral data into sub-intervals of the same width according to the characteristics of their value range.

Starting from the content and characteristics of the data, the steps of data mining consist of nine parts, including problem definition, data screening, data cleaning, data conversion, algorithm selection, data mining implementation process, pattern interpretation, knowledge evaluation and knowledge presentation. 1)

Problem Definition: Understand the data and business requirements, put forward the problem to be solved by data mining, determine the goal of data mining, and select the knowledge model of data mining.

The analysis is done by calculating the correlation coefficient between the data objects. Commonly used methods in the process of correlation analysis of data objects are Pearson correlation coefficient, Spearman rank correlation coefficient and coefficient of determination. 1)

Pearson correlation coefficient method

Pearson correlation coefficient is generally used to analyze the correlation between the behavioral data of data objects, and the calculation method is shown in formula (3). 3r=∑i=1n(xi−x¯)(yi−y¯)∑i=1n(xi−x¯)2∑i=1n(yi−y¯)2

Where the correlation coefficient r has a range of values and the different values of −1 ≤ r ≤ 1, r indicate the degree of correlation between the behavioral data of the data objects. 4{ r>0 Positive correlation r<0 Negative correlation |r|=0 Indicates that there is no linear relationship |r|=1 Indicates a perfectly linear relationship

Where the correlation coefficient |r| ranges from 0 < |r| < 1, different values of |r| indicate different degrees of linear correlation between the behavioral data of the data objects: 5{ r>0 Positive correlation r<0 Negative correlation |r|=0 Indicates that there is no linear relationship |r|=1 Indicates a perfectly linear relationship

K-Means algorithm is a clustering algorithm for the detection of differences in the division of data elements between the problem, it is first selected from the sample set of K start clustering target, and then according to the rules of the algorithm requires that the calculation of the distance between the sample data objects to be carried out, according to the results of the calculation of the analysis of the sample data objects grouped together, the iterative calculations until the center of the clustering and then no change, to get K clustering results.K-Means K-Means clustering algorithm is implemented as follows:

Input: set the sample set S has n data objects and the number of clusters K: 6S={x1,x2,⋯xn},K={c1,c2⋯,ck}

Output: K cluster that satisfies the minimum criteria of the criterion function.

The processing flow is as follows: 1)

Arbitrarily select K samples from the sample set S as the initial clustering center.

The average distance of all sample data points is (9): 8Meandist(S)=2n(n−1)×∑i≠j,j=1nd(xi,xj) 3)

Recalculate the mean (center object) of each sample that has been obtained.

where c_i is the center-of-mass point of a data object of the same class, and the center-of-mass point c_i is defined as in Equation (10): 10ci=1|Ci|∑xi=τixj

c_i denotes the center of the ind family and |Ci| is the number of data objects in class C_i. 5)

Ends and K clusters are obtained.

The flow of the above K-Means algorithm implementation is shown in Fig. 2.

Figure 2.

K-Means Algorithm Proce dure

1)

The concept of information entropy

Information entropy is a measure of the amount of information required to eliminate uncertainty, a concept used in information theory to measure the amount of information, often used as a quantitative indicator of the information content of a system, which can be further used as the objective of the optimization of the system equations or as a criterion for the selection of parameters. In the process of decision tree generation, entropy is used as a criterion for the classification of the optimal attributes of the samples. In this paper, we will briefly introduce the concept of entropy.

Suppose the sample data set s={x1,x2,⋯,xn}, its probability density pi=P[X=xi], its self-information of a sample data point can be expressed as I(xi)=log1pi, and the information entropy of the sample data set can be expressed as H(X)=∑ipilog1pj(i=1,2⋯,n), stipulating that 0 log(0) = 0. The information entropy is used to measure the amount of information, if the greater the uncertainty, the greater the amount of information, the greater the entropy: if the uncertainty is smaller, the smaller the amount of information, the smaller the entropy [26].

In Equation (5) m denotes the number of dimensions of the data object and n denotes the number of sample data.

Step2 calculates the weights of each attribute of the data object and the weight of the attribute value of the ith data object corresponding to the attribute of the jrd dimension [27]. Firstly, the data need to be processed to compress the data density range to the interval [0, 1], the process is shown in equation (12) equation: 12Mij=aij/∑i=1naij

Equation (12) in which a_ij represents the attribute value and M_ij represents the attribute value weight, where i = 1, 2, ⋯, m, j = 1, 2, ⋯, n.

Step3 The formula for calculating the entropy value of the jth dimensional attribute is shown in (13): 13Hj=−1lnn∑i=1nMijlnMij

When M_ij = 0, there is M_ij ln M_ij = 0. If for a given j value a_ij all are equal, then Mij=aij/∑i=1naij=1/n, at which point H_j takes an extremely large value.

Step4 Calculate the coefficient of variation for the jth dimensional attribute as shown in equation (14): 14qj=1−Hj

Where the coefficient of variation is q_j. For a given j, when H_j is smaller, q_j is larger and the attribute is more important, and when H_j is larger, q_j is smaller, then the attribute’s clustering role is smaller. When H_j = 1, q_j is zero, then the clustering role of the attribute is zero.

Step5 Calculate the attribute weights formula for the jth dimension as shown in (15): 15wj=qj/∑j=1mqj

Step6 Use the Euclidean distance formula to calculate the similarity between data objects, according to the definition can be obtained after the assignment of the Euclidean distance as shown in Equation (16): 16dw=(xi,xj)=∑p=1mwp(xip−xjp)2

In Eq. (16), w_p is the weight value of the attribute in the pnd dimension. It is equivalent to making the attribute values corresponding to the weights appropriately enlarged or reduced, so that attributes with large weights have a greater role in clustering and attributes with small weights have a smaller role in clustering.

Step7 adopts the standard deviation as the standard measurement function to find the assigned category objective value function as shown in Equation (17): 17σi=∑ai∈τjdw(ai,c(Cj))|Cj|−1

In formula (17), σ_i indicates the standard deviation of the assignment of class i; |Cj| is the number of data objects contained in C_j. As can be seen from the formula (17), the smaller the value of σ_i, indicating that the greater the similarity of the data objects within the class, the denser the data objects, the more the center of mass of the class where the data objects are located can reflect the classification decision surface. 4)

Optimization algorithm initial clustering center selection method

The selection of the initial clustering center of the traditional K-Means clustering algorithm is mostly based on empirical judgment of random selection, and the practice of randomly selecting the initial clustering center will lead to the instability of the algorithm, and may fall into the local optimal situation, which leads to the inaccuracy of the clustering results. The K-Means algorithm based on information entropy and density optimization studied in this paper is to measure the similarity of data points by weighting the Euclidean distance between two data points, and at the same time, in the selection of clustering centers by calculating its objective value function, according to the size of the value of the sort can be more accurate initial clustering centers. The process of the initial clustering center selection method of the optimization algorithm is as follows. (1)

Randomly select a specified number of k′(k′>k) initial clustering center data points;

It can be seen that the optimization algorithm is simple and fast in the process of selecting the initial clustering centers, and the clustering seed centers can be determined by sorting the size of the assigned objective value function values at one time, which is different from the random selection of traditional algorithms, and it can greatly reduce the chances of the initial clustering centers being divided into the same clusters.

The introduction of correlation objects is intended to uncover interesting connections between various types of student data, which can then be informative for both self-improvement and decision-making on the part of the student and the school. We introduce some concepts of association rules as follows. 1)

Item set

In the Wal-Mart example above, the sum of all items purchased by a customer is the itemset (e.g., all items sold at Wal-Mart form a set, which is the itemset), and each item is an item. This is represented by the formula shown below: 18I={i1,i2,i3,…in}

where I is denoted as the item set, which contains all the items i_j, and i_j denotes a specific item. 2)

Database

Again take the above Wal-Mart Supermarket as an example, the database is a list of each customer’s purchases as a base unit of data composition of the database, as shown in the following formula: 19D={t1,t2,t3,…tn}

where t_j is a purchase record which consists of a number of items in I, i.e., t_j ⊆ I is called a transaction, and all the transactions make up the database D. 3)

Rule

In the association rules, if X ⊆ I, Y ⊆ I, can be X ⇒ Y and X ≠ ∅, Y ≠ ∅, X ∩ Y = ∅, such as this type of association rules, X, Y are a subset of I, and there is no intersection between them, when X can be deduced from Y, X and Y are linked.

When we want to study the association rules between X and Y, we must first ensure that the frequency of X and Y in the database reaches a certain number, otherwise the small probability of occurrence is not enough to ensure the reliability of the association, as shown in the formula above, the formula for calculating the degree of support of X and Y is the number of occurrences of X and Y divided by the total number of all the matters that is, the degree of support of X and Y. 5)

Confidence: the probability of the appearance of one data after the appearance of another data, or the conditional probability of the data.

When the support of X, Y meets the minimum level of support we set, then we have to verify whether the correlation between them really exists, which will be assessed by the confidence level. The formula for confidence level is as follows: 21Confidence(X⇒Y)=P(XY)=P(XY)P(Y)

To summarize, the mining of association rules is mainly carried out by two steps, the first step is to find out the high-frequency item sets, and the second step is to generate association rules.

Assuming that {A,B,C} is a combination to be mined for association rules, there is one thing that we must ensure before mining the relationship between them, that is, the frequency of occurrence of combination {A,B,C} in all data sets must reach our preset minimum value, i.e., Support(A, B, C) ≥ P, where P denotes a certain probability proportion of occurrence or a certain number of occurrences. This ensures that the occurrence of the combination is not by chance, but accompanied by a certain percentage of probability of occurrence, the setting of P must be reasonable, if it is not possible to eliminate the chance time, the mining of association rules will be meaningless, adding to the amount of arithmetic.

The second step is to generate the association rules, this step is to ensure that the confidence of the combination, the same we will artificially select a minimum confidence to ensure that Confidence(A, B ⇒ C), thus proving that the occurrence of A, B will lead to the occurrence of C events, only through these two steps, can prove that the event for the rule associated events.

The core idea of Apriori algorithm is to derive K + 1 combinations from K combination, instead of the original idea of rule mining to find K + 1 combinations from the complete dataset that meets the minimum support, which relies on the theory that: if the K + 1 combinations are high-frequency combinations, the K subsets of the combinations must also be high-frequency subclasses, and the core process of Apriori algorithm: 1)

Connection step

Apriori algorithm is based on the a priori property of frequent itemsets, so we can obtain C_k by cross connecting L_k−1. Let the items in L_k−1 of two crosses are I_j, I_k, and they are arranged in dictionary order, when the items in L_k−1 are cross-connected, it is required that the first k-2 items of I_j, I_k are the same (Ij[1]=Ik[1]∧Ij[2]=Ik[2]∧…∧Ij[k−2]=Ik[k−2]), then I_j, I_k can be connected, in order to ensure that the result of the connection is not duplicated I_j[k − 1] < I_k[k − 1].

The improved calculation is: 23∑(Lk−1)1(Lk−1)|Lk−1||Ck|Ck1

The whole process of cropping is: check the subset in L_k−1 with the subset of all items in C_k respectively, and assign a value to that C_k item according to the matching situation, so it is divided into two cases for discussion as follows: 1)

When |Lk−1|−Ck1+1>Ck1, that is, when the matching process of the items in L_k−1 with the items in C_k, if the number of non-compliant items exceeds the number of subsets of each item in Ck1, we use the number of matches that are judged to be compliant to eliminate the non-compliant items in C_k in the pruning process.

The influence of gender on students’ performance has always attracted much attention, and related studies have shown that there are some differences between boys and girls in terms of intellectual development, comprehension, imagination, calculation, logic, memory and emotion at different times. The comparative analysis of male and female students’ academic performance in public and compulsory courses is shown in Table 1. Table 2 shows the statistics of achievement distribution.

It can be seen that, whether it is the average credit score of public courses or the average credit score of specialized courses, the overall situation of girls is better than that of boys, and the variance of girls’ results is smaller, indicating that the distribution of results is more stable than that of boys.

Figure 5 shows a comparison of the distribution of grades by gender. Where (a) is the performance in public courses and (b) is the performance in specialized courses. The yellow line indicates girls’ grades and the purple line indicates boys’ grades, and it is obvious that girls’ grades are better than boys’. Overall, girls have fewer failures and a greater proportion of high grades, while boys’ grades are mostly in the middle.

Figure 5.

Comparison of distribution of different gender scores

Poor diet, sleep, hygiene and other aspects of life, accumulated over time will become an impact on the physical and mental health of college students, and ultimately lead to the occurrence of disease and affect their studies. In this paper, we select the average monthly consumption amount, the average monthly consumption times, and the regular work and rest index calculated based on the average monthly early times and the average monthly late times (regular work and rest index = 60 points + average monthly early times - average monthly late times) to study and analyze its impact on students’ overall performance. To a certain extent, the amount of consumption and the number of consumption can reflect whether the students eat and drink reasonably in the cafeteria and supermarket in the process of campus life and study, whether there is a tendency to often stay in the dormitory and order takeout or go out off-campus; the number of times of waking up early and the number of times of returning home late can measure whether the students are regular in the process of campus life and study, whether there is a late return to the dormitory and sleep, resulting in low classroom attendance and poor learning results, and even whether there is a tendency to stay out late and violate the rules of the students’ grades. Even the number of late nights violates dormitory management rules and other regulations.

First, based on the students’ weighted average scores, the students were divided into three categories: those with scores higher than 80 were regarded as excellent; those with scores lower than 80 and higher than 60 were regarded as moderate; and those with scores lower than 60 were regarded as abnormal. The distribution of average monthly consumption amount, average monthly consumption frequency, and regular work and rest index of the three categories of students were studied and analyzed separately, and the results are shown in Figures 6, 7, and 8, respectively.

Figure 6.

Monthly average consumption

Figure 7.

Average number

Figure 8.

Regular sleep

Blue color indicates students with abnormal performance, yellow color indicates students with medium performance, and purple color indicates students with good performance. As can be seen from the figure, there are obvious differences between students with good, medium and abnormal performance in terms of the number of times of consumption, the amount of consumption, and the index of regular work and rest. Students with good grades and medium grades have more concentrated consumption and consumption amount, smaller fluctuation range, and the former is slightly higher than the latter; students with abnormal grades have lower consumption and consumption amount, and the change range is bigger. It means that students with better grades have more regular and reasonable consumption behavior, basically all three meals are consumed in the on-campus cafeteria and supermarket, and the behavior of traveling off-campus and ordering food from takeaways is less; students with poor grades obviously consume less in the on-campus cafeteria and supermarket, and most of the students may stay in the dormitory for a long time and rely on takeaways, and a small number of them may like to buy snacks and snacks instead of having meals in the right place, which leads to unhealthy behaviors such as consumption. A few students may have unhealthy behaviors such as preferring to buy snacks and not eating at the right time, resulting in significantly higher consumption and money consumption than average. In terms of students’ regular work and rest, the average number of times of waking up early is greater than the number of times of returning home late, and students with good habits of going to bed early and waking up early are obviously more inclined to have medium and good grades; while the frequency of staying up late and returning home late and the frequency of sleeping too much lead to low regular work and rest indexes, and their grades are more concentrated in the category of abnormality, which indicates that poor work and rest habits and consumption habits will to a certain degree lead to the disruption of the work and rest time, insufficient sleep time, unhealthy eating behavior, and thus unhealthy eating habits. This indicates that poor work and rest habits and consumption habits can, to a certain extent, lead to disturbed work schedules, insufficient sleep time, and unhealthy dietary behaviors, thus affecting the academic performance and physical and mental development of college students.

Mining social network relationships through the student card data can, to a certain extent, reflect the student’s personality and psychological problems, helping educators to more comprehensively understand the student characteristics of the information, identifying lonely students to focus on, verifying the actual life and learning situation of the students, and carrying out targeted guidance measures and interviews to promote healthy growth.

In order to facilitate the calculation, first of all, the time of consumption is expressed as a numerical form, and the place of consumption in the cafeteria supermarket is digitally coded, and the consumption data of 10 students in a certain month is randomly selected to calculate the matrix of co-occurrence times, as shown in Table 3.

Let the maximum value of the total number of co-occurrences among all students with other students be P, and the minimum value be Q. The total number of co-occurrences between each student and other students is standardized, and the total number of co-occurrences and the comprehensive weighted average grade are shown in Figure 9.

Figure 9.

Students Standardize the total number of credit CARDS

From the line graph, it can be seen that the trend of the total number of students’ standardized co-occurrences and the comprehensive grade has a certain isotropy. In order to verify the correlation between the total number of students’ standardized co-occurrences and the composite grade, a correlation analysis was performed.

The resulting density plots for the two variables are shown in Figure 10. The purple curve in the graph shows the distribution of the total number of times students have interacted with other students, while the red curve shows the distribution of the composite weighted average grade. It can be seen that the distribution of the two variables is significantly different, and does not indicate a strong correlation between the two, i.e., the number of times students co-occur with other students does not have a significant impact on the relationship between the student’s composite weighted average grade, after all, there are both solitary schoolmasters and schoolmasters who make a wide range of friendships, but in general, students who have a better interpersonal and social relationship are more likely to perform well in terms of their grades as well.

Figure 10.

The total number of times and the distribution of comprehensive results

Based on the above analysis of the characteristic indexes that have significant relevance and influence on students’ performance, a student with the student number 2020****107 in the dataset is randomly selected, and the various types of data on his campus data are subjected to the extraction of indexes, the analysis of characteristic behaviors, and the construction of the user portrait. The first grades are shown in Table 4.

The weighted average grade for professional courses is 85.71, and the weighted average grade for public courses is 84.95, which can be considered a good level of performance for this student.

Figure 11 shows the amount of money spent and the number of times spent by this student in each month. The average monthly consumption amount is 745.71 yuan, and the average number of consumption times is 75.86 times. The fluctuation of the consumption amount and number of times consumed in each month is not significant, and all of them are within the reasonable range, which means that the student’s consumption level is normal. Considering and comparing the real situation of the student, the judgment of the system in this paper is in line with reality.

Figure 11.

Monthly consumption and consumption

In conclusion, it can be concluded that the application of the system in this paper can realize the accurate construction of student profiles.

1)

Cluster analysis of students’ consumption habits

Using the information entropy and density optimization K-means algorithm based on information entropy and density optimization designed in the previous section, the consumption data of students are clustered. According to the evaluation guidelines of clustering algorithm, in order to get better clustering effect, the best clustering effect is obtained when the number of clusters is set to 5. We analyze the results of student classification after clustering, according to the actual situation of students’ consumption in the school, comparing the average value of each cluster with the average value of the overall indicators of students, and noting that H is the average value higher than the overall indicators of students, and L is the average value lower than the overall indicators of students. Table 5 displays the results of the cluster analysis.

It can be seen that the first type of students have a lower level of consumption per month, and the peak value of consumption in a single month is not high, but the number of times is more frequent, belonging to the group with a lower level of consumption. This type of students mostly belong to the more frugal type, is the economic conditions of the poorer students, it is recommended that the school and counseling teachers pay attention to the living conditions of this type of students, the identification of poor students and financial assistance and other work can be considered from this type of students to choose. The second type of students have a medium level of consumption per month, the average consumption per month is not high, but the maximum value of consumption in a single month is high and the number of times of consumption is also low. This type of student should eat irregularly in the school cafeteria and usually prefer to eat off campus or order takeout. The third type of student has a moderately high monthly consumption level, a high frequency of monthly consumption, and a low variability in the amount of consumption. This indicates that this type of student often eats in the school cafeteria and has stable consumption, which is in line with the normal eating pattern of more students at school.

The fourth type of students had the highest average consumption per month, frequent consumption and higher maximum value of consumption, indicating that this type of students is a high-consumption group in the cafeteria, and that students eat in the cafeteria on a more regular basis, with higher requirements for the quality of meals and larger meal sizes.

The average monthly consumption of students in the fifth category is lower and less frequent, and the maximum value of consumption is also not high. Students in this category do not frequently eat in the cafeteria, and are more likely to eat off-campus frequently and consume more off-campus. Schools and counselors, etc. Should also pay attention to food safety and personal safety, etc. of such students who often go off-campus.

In the parameter setting of the Apriori algorithm model, five types of students’ consumption habits, three types of living habits, four types of study habits and three types of comprehensive qualities are set as the pre and post variables of the association rules as shown in Table 8, Table 9 and Table 10 below.

Setting the support degree as 10% and confidence degree as 80%, the mining of association rules is carried out, and a total of 24 association rules are mined to get the association network diagram as shown in Fig. 12, according to the goal of this paper to be researched, the redundant rules are eliminated and merged, and the association rules with the degree of enhancement of more than 1 are selected, among which the later items of the association rules for the comprehensive quality of the students are shown in Table 11.

Figure 12.

Associated network

The association rules mined were life habit type 1, study habit type 1 → good general quality (S=10.317%,C=81.362%); Habits of life type 3 → average overall quality (S=53.124%, C=84.194%); Consumption Habit Type 5, Life Habit Type 3 → General Quality (S=11.956%, C=83.251%); Student Habit Type 4, Life Habits Type 1 → Average overall quality (S=23.705%, C=80.578%); Life Habit Type 2 → Poor Comprehensive Quality Type (S=10.272%, C=82.566%) The degree of enhancement of these five association rules is greater than 1, which indicates that the occurrence of the antecedent event of the association rule favors the occurrence of the latter event.

The first rule reflects that when the student’s life habit is to get up early, go to the cafeteria and eat regularly and exercise regularly, and the study habit is to have a high rate of attendance in class, borrow more books, go to the library regularly and study for a longer period of time, then the student has a probability of 81.343% to have a better quality of overall performance.

The second rule reflects that when a student’s living habits are often early, but diet is not regular and does not often exercise, then the student has a probability of 84.068% of the overall quality of the average.

The third rule reflects that when the student’s spending habits are low total monthly spending, less frequent spending, and living habits are often early, but not regular diet and not regular exercise, then the student has 83.246% probability of having average overall quality.

The fourth rule reflects that when a student’s study habits are average class attendance, borrowing fewer books, going to the library less often and spending less time studying, and living habits are often early, eating more regularly, and exercising regularly, the student has 80.366% likelihood of having average overall quality.

The fifth rule states that if the student’s life habits are not early, eat irregularly, and exercise infrequently, they have an 82.423% probability of having poor overall quality.

To sum up, it can be seen that students’ behavioral habits during the school period are closely related to their comprehensive quality, and this paper obtains some laws affecting the development of students’ comprehensive quality by digging the relationship between various indicators of students’ behavioral characteristics and their comprehensive quality. In the students’ living habits, learning habits and consumption habits will have an impact on the development of students’ quality, good learning habits and living habits are conducive to students’ rational planning of their time, better learning professional knowledge, improve the level of practice, and disperse innovative thinking, and comprehensively enhance the overall quality of students.