Research on Slow Travel Consumer Behavioral Feature Extraction and Decision Support Based on Intelligent Data Analysis

Acceleration is the core feature of modernity. Against the backdrop of the accelerated development of informationization, digitization and intelligence, the impact of the logic of social time on the pace of life has become more and more obvious, and the so-called “time drought” has even left people with no time to think about what they need and want. To a large extent, the social acceleration under the view of capital has hindered the realization of the “good life” of social members, and its serious impact on the physical and mental health of individuals has attracted great attention from all social actors [1–2]. In the face of the demand for deceleration stimulated by all kinds of life suppression, business models with the theme of “slow” have accelerated in recent years, and “slow tourism” is also included [3–4]. The “slow” that tourists feel is equivalent to quality time, a meaningful experience in harmony with cultural and natural environments, which includes both physical and mental deceleration in enjoying time, as well as interactions with others and nature [5–7]. The “slowness” associated with deceleration is a way of being in relation to “speed”, and it is a qualitative rhythm of experience between consumers and the world based on sensory and emotional connections [8–9]. Consumer slowdown is both a behavior and a perception, in which changing consumption behavior is the inevitable way to achieve slowdown, and the perception of slowing down time experience is the ultimate goal of slowdown [10–11].

Taken together, the spread of “slow” behavior not only reflects people’s resistance to the “cult of speed” in the fast-paced society, but also reflects people’s expectation of alternative and new life [12]. These new expectations related to “slowness” not only provide people with an “oasis of deceleration” against acceleration, but also make “slow” consumption an effective way for members of society to change their pace of life and achieve self-deceleration [13–15].

Literature [16] shows that mass tourism has always been regarded as the main form of tourism, but with the change of tourism market demand, independent or selective tourism that can provide tourists with completely different experiences has begun to replace the position of mass tourism in the tourism market, and “slow tourism” has begun to become one of the new trends in contemporary tourism. Literature [17] used the push and pull theory to analyze the driving factors affecting the behavioral decisions of slow tourism tourists, and found that the push factors that motivate people to choose the form of slow tourism include six factors such as “relaxation” and “self-reflection and discovery”, and the pull factors are the study provides a valuable reference for tourism destination marketers to design tourism products and services. Literature [18] explored users’ attitudes, travel behaviors, and willingness to pay in sustainable slow tourism, which centers its activities on close contact with the culture and natural environment of the destination and combines with fast transportation to achieve the sustainability of mass tourism, which moreover increases users’ willingness to pay. Literature [19] analyzes users’ behavioral willingness to participate in slow tourism by constructing a slow tourism research model and analyzing the three aspects of slow travel mode, tourism experience and environmental awareness, and the study shows that the quality of transportation mode and tourism experience are the determinants of tourism users’ participation in slow tourism. Literature [20] used the theory of planned behavior as a framework to examine the individual factors, tourist satisfaction and revisit intention of slow tourism in developed and developing countries, and used it to construct an extended model to further understand the decision-making process of slow tourism tourists. Literature [21] takes slow tourism destination city attributes as a perspective to study its influence on users’ behavioral intentions, and the structural analysis proves that attributes such as sense of belonging, psychological well-being, and intention are important factors to promote the positive behaviors of tourism users, which provides a certain theoretical foundation for tourism marketers.

Research based on data cleaning, data de-duplication, and data noise reduction to establish the key data set for slow tourism consumption. Data mining technology is being used in big data technology to set the clustering method for object key features based on big data. And the ant colony algorithm is invoked to construct the slow tourism consumer behavioral feature extraction process, fusing to extract the subject/object key features. Starting from the extracted behavioral features, the interest feature extraction model is established. By constructing the interest degree matrix, N slow tourism attractions with high interest degree values are selected and recommended to consumers to provide decision support.

2

Decision support based on slow tourism consumer behavioral characteristics

2.1

Slow Travel Consumer Behavior Feature Extraction Methods

2.1.1

Key dataset setup for slow tourism consumer subjects

Collect slow tourism consumer behavior data and set up the key dataset of consumer subjects. Since the original data contains a large amount of duplicated data or residual data, it is necessary to carry out preliminary processing of this dataset. The study sets the key dataset processing part as three parts: data cleaning, data sampling, and feature preprocessing. The operation process is defined as follows: 1)

Data Cleaning

The collected consumer behavior data contains a large number of outliers, in the analysis process, this part of the data will have an impact on the analysis process, for this reason, it is necessary to filter them, and through the maximum and minimum value calculation to remove abnormal values in the data. For the detected outliers, such as the number of small cases, they can be ignored. When the total number of outliers is high, the average value can be used to replace this data or directly remove this part of the data [22].

2)

Data de-weighting

The collected raw data contains part of the same data volume, in order to improve data quality and reduce the difficulty of data processing. Use the similarity function to calculate the data to obtain the similarity value results. Set the corresponding threshold value before calculation. If the data calculation results exceed the threshold value, it means that this data is redundant and needs to be eliminated.

3)

Data noise reduction

In the study, regression is used to complete the data fitting, and the processed data is used to replace the original data, so as to achieve data noise reduction. In the process of this operation, the data needs to be visualized and analyzed for trends to complete the noise reduction.

2.1.2

Fusion to extract key subject/object features

Due to the large number of slow tourism consumer behavior data features and the correlation between most of the categories, only the application of big data technology can not be fused to analyze the object features and the subject features to complete the extraction of all the features, therefore, the ant colony algorithm is invoked in the study to realize the feature fusion and extraction on the basis of the results of the classification of the subject data as well as the results of the mining of the object data [23]. Considering the feature data space as a Q*Q two-dimensional space, an ant can retrieve the data within its surrounding O*O range, and the data similarity of two data features in this region can be expressed as: 1 $f (Y_{i}) = \max {0, \frac{1}{w^{2}} \sum [1 - \frac{d (Y_{i}, Y_{j})}{ω}]}$ where d(Y_i,Y_j ) denotes the distance between the two data categories. Y_i denotes the data category 1. Y_j denotes the data category 2. ω denotes the parameter that measures the similarity of the behavioral data categories. w² denotes the data probing range. Based on this formula, determining the definition of the ant retrieval process to retrieve the next object, there is: 2 $E (Z_{i}) = (\frac{k_{1}^{'}}{k_{1}^{'} + f (Z_{i})})$ 3 $E (Z_{i}) = {\begin{array}{l} 2 f (Z_{i}) \\ f (Z_{i}) < k_{2}^{'} \\ 1 f (Z_{i}) > k_{2}^{'} \end{array}$

Where, g(a) denotes the information entropy of the radius region of the data category. $h_{i}^{'} (x)$ denotes the number of data objects in the data set. Applying this method further reduces the time-consuming data clustering process and improves the computational efficiency. Afterwards, the content is integrated to obtain the following slow travel consumer behavior feature extraction process.

The detailed setting of the slow tourism consumer behavior feature extraction process has: 1)

Apply the data clustering method to set the rule set of consumer behavior data and display it as Result = ε_i M_i.

2)

Calculate the minimum confidence level of subject feature data and object feature data set according to the consumer behavior data rule set combined with the current feature extraction method, and output the corresponding association rules [24].

3)

Calculate the support degree of the collected data set for each category.

4)

If the quotient of the support degree of the data set as well as its non-empty subset is not less than the minimum confidence degree, output this data set, which is the slow tourism consumer behavioral characteristics for analyzing the preferences and needs of slow travelers in the travel process.

Organize the above contents, connect them in an orderly manner and apply them to the current process of slow tourism consumer behavioral feature extraction, so as to complete the process of applying big data technology in the slow tourism consumer behavioral feature extraction, and the method proposed in the paper is set to be completed.

2.2

Decision Support Based on Slow Travel Consumer Interests

2.2.1

Slow Travel Consumer Interest Feature Extraction Model

This subsection will give a model for extracting the interest features of slow travel consumers based on the slow travel consumer behavioral features extracted above, starting with some definitions.

Definition 1: If a slow traveler has visited the attraction at t₁ moment i, the probability of slow traveler visiting the attraction again at t₂ moments is denoted by α(0 ≤ α ≤ 1), and the probability of slow traveler forgetting the attraction is denoted by β, β = 1 – α and 0 ≤ β ≤ 1, β are called forgetting factors.

Definition 2: If the time of the slow traveler’s k nd visit to the attraction i of interest to the slow traveler is denoted as t_i,k, then the sum of the time of the slow traveler’s k visits is T referred to as the total time of the slow traveler’s visit to the attraction i.

Definition 3: If n out of the k visits of a slow traveler to an attraction i are within the specified time range, i.e., t_min ≤ t_i,j ≤ t_max, t_min, and t_max are the thresholds for the minimum and maximum visit times of the slow traveler to the attraction, then the sum of these n visit times is T′ called the effective visit time of the slow traveler to the attraction i.

Definition 4: The number of times a slow traveler consumer visits a certain attraction i is denoted as k, then it is said k to be the total frequency of visits to the attraction i by the slow traveler consumer, denoted by f.

Definition 5: Let the current browsing time of a slow traveling consumer for attraction i be t_n, and the time of the last browsing of the attraction be t_l. If 0 ≤ t_n – t_l < t_j, the slow traveling consumer’s last browsing of the attraction is considered to be effective, which is called the effective browsing frequency, and is denoted by e_j.

For the understanding of slow tourism consumers’ interest characteristics, this paper considers that the interest degree of slow tourism consumers in a particular attraction should follow the following principles:

Principle 1: If the total browsing time of slow travel consumers for attraction i, j is the same, and if the total browsing frequency f_i of slow travel consumers for attraction i is greater than the total browsing frequency f_j for attraction j, then the interest level of slow travel consumers for attraction i is higher than that for attraction j, all other factors being equal. Expressed as a mathematical relationship: 4 $d_{i} > d_{j} for f_{i} > f_{j}$

Principle 2: If the effective browsing time t_i of slow travel consumers for attraction i is greater than the effective browsing time t_j of attraction j in the same period of time, other factors remain unchanged, the interest level of slow travel consumers for attraction i is considered to be higher than the interest level for attraction j. Expressed as a mathematical relationship equation: 5 $d_{i} > d_{j} for t_{i} > t_{j}$

Principle 3: If slow travel consumers have the same total browsing frequency for Attractions i, and if the total browsing time T_i for Attractions i is greater than the total browsing time T_i for Attractions j, while other factors remain unchanged, it is considered that the interest level of slow travel consumers in Attractions i is higher than that in Attractions j. 6 $d_{i} > d_{j} for T_{i} > T_{j}$

Principle 4: If a slow traveler does not revisit an attraction of interest for more than a period of time k, the slow traveler’s interest in the attraction is considered to have changed little. Let the current time of the system be t and the last time the slow travel consumer visited the attraction be t′. Expressed as a mathematical relationship: 7 $d_{i, t} \approx d_{i, t}, for t - t' > k$

In order to satisfy the above four principles, the following parameters are elicited: system parameter recent weighting factor α, system parameter time of interest coefficient ρ_i, system parameter k, system time difference t_α, current effective browsing time of the slow travel consumer f_(t). where: 8 $f_{(t)} = (1 - β) t + β$

Based on the above parameters, the method proposed in this paper for calculating the interest level of a slow traveling consumer in an attraction i is: 9 $d_{(i)} = {\begin{matrix} (1 - α) f^{T / p_{t}} + α e_{f}^{f (t) / p_{t}} & t_{a} \leq k \\ d_{(i)} + α e_{f}^{f (t) / p_{t}} & t_{a} > k \end{matrix}$

Theorem 1: The calculation method of d_i in Eq. (9) satisfies Principles 1 to 4.

2.2.2

Constructing the Slow Travel Consumer Interest Level Matrix

This paper utilizes the Pearson similarity coefficient to initially assess slow travel consumers’ interest in unvisited slow travel attractions as a way to increase the number of slow travel attractions of common interest to slow travel consumers.

Let the set of slow tourism attractions that have not been viewed by slow tourism consumer u in the spatial set of slow tourism attractions U_a,j be denoted by N_a, and the set of slow tourism attractions that have been viewed be denoted by I_a. Then, N_a = U_a,j =– I_a, the method of predicting the degree of interest of a slow tourism consumer in a slow tourism attraction ρ for any slow tourism attraction p ∈ N_a is as follows: 1)

Find the interest degree values of all slow tourism consumers who have viewed slow tourism attractions i and j in slow tourism attraction set I_u, and calculate the similarity between slow tourism attractions i and j using the traditional similarity measure.

2)

Select the k slow tourism attractions with high similarity to form the slow tourism attraction similarity set M_k = {I₁,I₂,…I_k}, where p ∉ M_p, and make sim(p, I₁) > sim(p,I₂),…> sin(p,I_k ).

3)

In this paper, we use equation (10) to predict the interest degree of slow tourism consumers in slow tourism attractions that have not been visited: 10 $p_{j} = \frac{\sum_{k \in M_{k}} \sin (j, k) \times D_{k}}{\sum_{k \in M_{k}} \sin (j, k)}$

After processing by the above method, the interest degree of slow tourism consumer u for any slow tourism attraction i is obtained as: 11 $D_{u, i} = {\begin{matrix} d_{i} & if user u browse item i \\ p_{i} & if user u not browse item i \end{matrix}$

On this basis, we construct n×k a matrix of slow tourism consumers’ u interest in slow tourism attractions j, n rows represent n slow tourism consumers, k columns represent the selected k slow tourism attractions, and D_i.,j denotes the slow tourism consumers’ u interest value in slow tourism attractions k, and the slow tourism consumers’ interest matrix is as follows: 12 $[\begin{matrix} D_{1, 1} & D_{1, 2} & D_{1, 3} & \dots & D_{1, k} \\ D_{2, 1} & D_{2, 2} & D_{2, 3} & \dots & D_{2, k} \\ D_{3, 1} & D_{3, 2} & D_{3, 3} & \dots & D_{3, k} \\ ⋮ & ⋮ & ⋮ & ⋮ \\ D_{n, 1} & D_{n, 2} & D_{n, 3} & \dots & D_{n, k} \end{matrix}]$

2.2.3

Generate decision support

In order to produce accurate decision support results, the slow tourism consumers’ interest degree values for unviewed slow tourism attractions are utilized and the top-V slow tourism attractions with higher interest degree values are selected to be recommended to slow tourism consumers.

Based on the established interest degree matrix of slow tourism consumers for slow tourism attractions, Pearson correlation measure is used to find the set of neighbors with similar interests to slow tourism consumers, on the basis of which the interest degree of slow tourism consumers for unviewed slow tourism attractions is predicted. The similarity measure is as follows: 13 $\sin (i, j) = \frac{\sum_{c \in I_{i, j}} (D_{i, c} - \bar{D_{i}}) (D_{j, c} - \bar{D_{j}})}{\sqrt{\sum_{c \in I_{i, j}} {(D_{i, c} - D_{i})}^{2}} \sqrt{\sum_{c \in I_{i, j}} {(D_{j, c} - D_{j})}^{2}}}$ Where: I_i,j denotes the set of slow tourism attractions that slow tourism consumers i and j are jointly interested in D_i,c denotes the interest level of slow tourism consumer i in slow tourism attraction c, and ${\bar{D}}_{i}$ denotes the average of the interest level of slow tourism consumer i in all slow tourism attractions.

For slow tourism consumer u, find the set of slow tourism consumers U = {u₁,u₂,⋯,u_k} with similar interests to it in the entire space of slow tourism consumers such that u ∉ U, and sin(u, u₁) > sin(u, u₂) > ⋯ > sin(u,u_k).

The method of inferring the slow tourism consumer’s interest degree value for an unvisited slow tourism attraction based on the set of slow tourism consumers is as follows: 14 $P d_{u, i} = \bar{D_{u}} + \frac{\sum_{k \in U} \sin (u, k) \times (D_{k, i} - \bar{D_{k}})}{\sum_{k \in U} \sin (u, k)}$

Where: $\bar{D_{a}}$ and ${\bar{D}}_{k}$ denote the average degree of interest in slow tourism attractions between slow tourism consumers u and neighboring slow tourism consumers k, sin(u, k) denotes the degree of similarity of interest between slow tourism consumers u and neighboring slow tourism consumers k, and D_k,i denotes the degree of interest in slow tourism attractions i by neighboring slow tourism consumers k.

2.2.4

Metrics

The study uses the mean absolute error (MAE) to verify the decision support effect, which measures the accuracy of the method’s prediction by the deviation between the predicted consumer’s interest in a slow tourism attraction and the consumer’s true interest in the attraction, and the smaller the value of MAE, the more accurate the prediction, and the more accurate the decision support recommendation [25]. Let the value of slow tourism consumers’ interest in k slow tourism attraction predicted by the algorithm is {d₁,d₂,⋯,d_k} respectively, and the value of consumers’ real interest is {r₁,r₂,⋯,r_k}, then the mean absolute deviation is described as: 15 $M A E = \frac{\sum_{i = 1}^{k} | d_{i} - r_{i} |}{k}$

3

Empirical analysis

3.1

Analysis of the effect of behavioral feature extraction

The experiment selects the slow travel consumer behavior datasets RESSET and TOUR as the data source, and verifies the time-consuming and extraction accuracy of the behavioral feature extraction method, the base method, and the neural network method in this paper, respectively. The time-consuming and accuracy comparison of the three behavioral feature extraction methods on the RESSET dataset is shown in Figure 1. On the RESSET dataset, the time consumed for behavioral feature extraction is reduced from 45s for the base method to 31s using the method in this paper, which has a significant effect on the overall speed improvement. On the RESSET dataset, compared to the 79.65% accuracy of the neural network method, 98.78% accuracy is achieved using this paper’s method.

A comparison of the time-consumption and accuracy of the three behavioral feature extraction methods on the TOUR dataset is shown in Figure 2. The time consumption is reduced from 33s using the neural network method to 26s in this paper’s method, and the recognition accuracy is improved from 65.85% in the base method to 97.22% in this paper’s method.

Taken together, the slow travel consumer behavior feature extraction based on this paper’s method on two different datasets has the highest accuracy and the shortest running time.

3.2

Analysis of the effectiveness of decision support

In order to verify the decision support effect of the method in this paper, the TOUR dataset is selected as the experimental data, and the final experiment is randomly divided into the training dataset and test dataset according to the proportion of 80% and 20% of the dataset for analysis and comparison.

In order to select the appropriate number of nearest neighbors to achieve a better decision support effect when tuning the parameters, this experiment is done to analyze the MAE value of different number of nearest neighbors. The interval of the value of the nearest neighbor users is set to [5, 50], and the increment is 5, and then the MAE is calculated, and the experimental results are shown in Fig. 3. The increasing number of nearest neighbor users, the MAE value will gradually decrease, when the value of nearest neighbor users is 30, the MAE tends to stabilize, which is better, so the nearest neighbor users are selected as 30 in the later tuning parameters.

The traditional collaborative filtering algorithm based on cosine similarity (UCB-CF), collaborative filtering algorithm based on Pearson’s similarity (UPB-CF), collaborative filtering algorithm based on user’s characteristics (CFA-UC) and this paper’s method are selected for comparison, and the nearest-neighbors interval of the four methods are all [5, 50] with an increment of 5. The result of MAE is shown in Fig. 4. The MAE of all four recommendation algorithms decreases gradually, and the mean values of MAE of UCB-CF, UPB-CF, CFA-UC and this paper’s method in the number of nearest neighbors [5, 50] are 0.889, 0.861, 0.831, 0.797, respectively, and this paper’s method obtains a smaller average absolute error than the other three methods, with the best decision support effect.

3.3

Analysis of satisfaction with decision support

The study conducted a questionnaire survey with four and five items under the dimensions of perceived usefulness and perceived intrusiveness, respectively, and analyzed the satisfaction with decision support for slow tourism attractions based on the methodology of this paper based on the results of the questionnaire measurement scale.

3.3.1

Descriptive statistical analysis of samples

The effective sample size of this experiment is 321 people, and the basic situation of the experimental sample is shown in Table 1. In terms of gender, 61.68% and 38.32% of the respondents were female and male, respectively. In terms of age structure, the youth group aged between 18 and 30 years old is the majority, with a proportion of 48.60%, followed by middle-aged people aged between 31 and 40 years old, occupying a proportion of 34.27%. In terms of educational attainment, the proportion of subjects with bachelor’s degree accounted for the majority, with a proportion as high as 62.31%, followed by master’s degree and above, with a proportion of 16.82%, indicating that the overall educational attainment and quality of the subjects were high. About 19% of the subjects had an average monthly income of more than 10,000 yuan, and the overall income level of the subjects was high.

Table 1.

The basic situation of the experimental sample

Sample characteristics	Subcategory	Frequency	Percentage
Gender	Male	123	38.32%
Gender	Female	198	61.68%
Age	18-30	156	48.60%
	31-40	110	34.27%
	41-50	33	10.28%
	51-60	22	6.85%
Education	Secondary school	17	5.30%
	Specialty	50	15.58%
	Undergraduate	200	62.31%
	Master’s degree	54	16.82%
Monthly income interval (Yuan)	≥2000	30	9.35%
	2001-4000	45	14.02%
	4001-6000	57	17.76%
	6001-8000	54	16.82%
	8001-10000	74	23.05%
	≤10000	61	19.00%

3.3.2

Scale reliability analysis

Reliability test is used to test the consistency and stability of the variables, SPSS 21.0 was used to analyze the reliability test of the variables, and the calculated Cronbach’s Alpha values are shown in Table 2, and the value of Cronbach’s Alpha for each variable is above 0.8, which indicates that the questionnaire measurement scale has good reliability.

Table 2.

Reliability Test of Questionnaire

Variable name	Cronbach’s Alpha	Term number
Perceptual usefulness	0.865	4
Perceptual intrusion	0.896	5

3.3.3

Scale validity analysis

The validity test is about how well the chosen measurement tool can accurately measure and reflect each variable. The results of the validity test analysis using SPSS 21.0 are shown in Table 3. The KMO values of perceived usefulness and perceived intrusiveness are 0.833 and 0.845 respectively, and the significance level is less than 0.001. This indicates that the variables measured by the selected measurement tools are valid and can be analyzed in the upcoming tests.

Table 3.

Validity Test of Questionnaire

Variable name	KMO	Approximate card square test	p
Perceptual usefulness	0.833	633.57	0.000
Perceptual intrusion	0.845	641.52	0.000

3.3.4

Experimental results

The result statistics of decision support satisfaction on perceived usefulness and perceived intrusiveness are shown in Table 4. ANOVA was conducted with perceived usefulness as the dependent variable and decision support satisfaction as the grouping variable. The difference between decision support satisfaction and dissatisfaction is significant with an F-value of 14.32 and a p-value of less than 0.01. The number of people in the decision support satisfaction group is 231 more than the number of people in the decision support dissatisfaction group, which indicates that the decision support under the methodology of this paper has a high level of perceived usefulness. ANOVA was conducted with perceived intrusiveness as the dependent variable and decision support satisfaction as the grouping variable. The difference between the two groups was significant (F=17.25, p<0.01). The number of people in the decision support dissatisfaction group was 49, indicating that the number of subjects who were satisfied with the decision support under the methodology of this paper was predominant and fewer subjects had perceived intrusiveness in decision making.

Table 4.

Decision support satisfaction analysis

Test variable	Decision support satisfaction	Descriptive statistics			ANOVA test
Test variable	Decision support satisfaction	N	M	SD	F	p
Perceptual usefulness	Satisfaction	276	5.632	1.236	14.32	0.006
Perceptual usefulness	Discontent	45	5.415	0.698	14.32	0.006
Perceptual intrusion	Satisfaction	272	3.268	1.362	17.25	0.005
Perceptual intrusion	Discontent	49	2.514	0.697	17.25	0.005

4

Conclusion

The study invokes the ant colony algorithm to construct the behavioral features extracted from slow travel consumers, and constructs an interest matrix based on the extracted behavioral features to provide decision support. On two different datasets, RESSET and TOUR, the accuracy of slow travel consumer behavioral feature extraction based on this paper’s method is 98.78% and 97.22%, and the running time is 31s and 26s, respectively, and the average absolute error obtained by this paper’s method is smaller than that of the three methods of UCB-CF, UPB-CF and CFA-UC. It shows that the decision support provided by this paper’s method is the best. The number of people satisfied with the decision support for slow tourist attractions under the dimensions of perceived usefulness and perceived intrusiveness is 276 and 272, respectively, which accounts for a higher percentage, and the p-value between the decision support satisfaction and dissatisfaction is less than 0.01, which is a significant difference. It shows that the satisfaction with decision support formed under the method of this paper is high.

Lingua:: Inglese

Frequenza di pubblicazione:: 1 volte all'anno
Argomenti della rivista:: Scienze biologiche, Scienze della vita, altro, Matematica, Matematica applicata, Matematica generale, Fisica, Fisica, altro

Feed RSS della rivista

Research on Slow Travel Consumer Behavioral Feature Extraction and Decision Support Based on Intelligent Data Analysis

Jing Wang

Pubblicato online: 19 mar 2025

Ricevuto: 13 nov 2024

Accettato: 22 feb 2025

DOI: https://doi.org/10.2478/amns-2025-0376

Parole chiaveAnt colony algorithm, Decision support, Behavioral feature extraction, Slow travel consumers

© 2025 Jing Wang, published by Sciendo

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

Parole chiave
Ant colony algorithm, Decision support, Behavioral feature extraction, Slow travel consumers