Clustering and chameleon group optimization-based whole house customized product personalized recommendation method

Abstract

The invention discloses a personalized recommendation method based on clustering and chameleon group optimization in a whole house customized product, which comprises the following steps: firstly, introducing a golden sine strategy in a chameleon group algorithm prey searching stage to enhance the global exploration capability of the chameleon group algorithm in the early stage of algorithm iteration; the influence of insufficient space exploration capability on the chameleon group algorithm is reduced, the global exploration capability and the convergence speed of the chameleon group algorithm are improved, and the improved algorithm is called a GSCSA (Golden Sine Sine-based chameleon group algorithm); secondly, the GSCSA is combined with a K-means clustering algorithm, so that an optimal user similar cluster is determined; and finally, combining the similar user cluster with a collaborative filtering algorithm to form an optimized GSCSA-KCF method. The method aims at providing more accurate personalized whole-house customized product recommendation for a user by using historical records of similar users when the user is subjected to whole-house customization.

Description

Technical Field

[0001] The present invention relates to the fields of recommendation algorithms and software engineering, and in particular, to a personalized recommendation method based on clustering and chameleon swarm optimization in whole-house customization products. Background Art

[0002] In modern society, with the development of information technology and the Internet, the demand for personalized customization has gradually emerged, and whole-house customization has become a popular trend in the home furnishing industry. Whether it is customizing product styles, space layouts, or material selections, consumers hope to create a unique living environment through personalized customization solutions.

[0003] In this context, recommendation algorithms have been widely applied in the field of whole-house customization, greatly improving the user experience by recommending customized product combinations and design solutions that meet the personalized needs of users. However, with the diversification of users' personalized needs and the increase in the types of customizable products available, the recommendation system faces many challenges.

[0004] Among them, how to recommend better combinations of customized products to better meet the personalized customization needs of users has become a key issue in the research of whole-house customization recommendation systems. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a personalized recommendation method based on clustering and chameleon swarm optimization in whole-house customization products. This detection method can enhance the global exploration ability of the chameleon swarm algorithm in the early stage of algorithm iteration, reduce the impact of insufficient space exploration ability on the chameleon swarm algorithm, improve the global exploration ability and convergence speed of the chameleon swarm algorithm, further obtain user similarity clusters, recommend users to the target user in its similarity cluster, and give the target user more accurate whole-house customization product recommendation information.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0007] A personalized recommendation method for whole-house customization products based on clustering and chameleon swarm optimization, comprising the following steps:

[0008] S1. Establish a recommendation data set and switch the recommendation data set into a training set and a test set;

[0009] S2. Initialize the chameleon population and the number of clusters K;

[0010] S3. Use the chameleon swarm algorithm to obtain K optimal similar user clusters;

[0011] S4. Use the collaborative filtering algorithm in the test set to score products in the similar user clusters of the user to be recommended;

[0012] S5. Sort the rated products of the similar user cluster corresponding to the user to be recommended according to the average score, and select the top N products with higher scores to form a personalized customized product list for product recommendation to the user.

[0013] Step S2 includes:

[0014] Initialize the clustering center and the number of clusters K of the similar user cluster, and use the precision of the collaborative filtering recommendation algorithm to evaluate the clustering quality of each similar user cluster.

[0015] In step S2, the initialization of the clustering center includes: initialize using the following formula:

[0016]

[0017] where is the initial position of the clustering center represented by the j-th chameleon, up j and low j represent the maximum and minimum boundaries of the accessible search area of the clustering center represented by the j-th chameleon, and r is a random number between 0 and 1.

[0018] In step S3, it includes:

[0019] S31. Update the clustering center of the similar user cluster using the improved chameleon search strategy;

[0020] S32. Update the clustering center of the similar user cluster using the chameleon hunting strategy and evaluate the clustering quality of this similar user cluster;

[0021] S33. Determine whether the iteration number threshold is reached. If not, feedback to S31; if so, use the position of the best chameleon individual in the current population as the K optimal similar user clusters.

[0022] The precision of the collaborative filtering recommendation algorithm includes:

[0023]

[0024] where R(u) represents the set of N products recommended to user u, and T(u) represents the set of products that user u likes.

[0025] The improved chameleon search strategy includes:

[0026]

[0027] where represents the position of the clustering center represented by the j-th chameleon at the t-th iteration; Gb t represents the best clustering center position represented by the optimal individual in the chameleon group up to the t-th iteration; up j and lowj represents the search space boundary of the clustering center represented by the j-th chameleon; p1 is a positive number that controls the exploration ability of the algorithm, r is a random number within [0, 1], represents the probability that the chameleon detects prey, the value of sgn(rand - 0.5) is +1 or -1, indicating the direction of exploration and exploitation; γ, α, and β are three constant values that adjust the exploration and exploitation abilities; r sine1 and r sine2 are random numbers between [0, 2π] and [0, π] respectively, representing the moving distance and direction of the next-generation individual; x1 = aτ + b(1 - τ) and x2 = a(1 - τ) + bτ, where a and b are -π and π respectively, and τ is the golden ratio t and T represent the current iteration number and the maximum iteration number respectively.

[0028] The hunting strategy of the chameleon includes:

[0029]

[0030] Among them, represents the speed of the chameleon's tongue when hunting the clustering center, a = 2590*(1 - e -log t ) represents the acceleration of the chameleon's tongue popping out.

[0031] The process of the user to be recommended scoring unrated products in the same cluster includes:

[0032]

[0033] Among them, represents the predicted score of user u for product i, N(u) represents the set of products with the historical scores of user u, represents the cosine similarity between user u and user v in the same cluster, I uv represents the set of products jointly rated by user u and user v in the same cluster, I u and I v represent the sets of products with positive feedback made by user u and user v in the same cluster in history, r ui and r vi represent the scores given by user u and user v in the same cluster for product i. Finally, the top N products with the highest scores among the target users are selected to form the personalized whole-house customization product list for user u.

[0034] The advantages of the present invention are as follows: The golden sine strategy is introduced in the prey search stage of the chameleon swarm algorithm, the position search equation of the chameleon swarm in this stage is improved, the global exploration ability of the chameleon swarm algorithm in the early stage of algorithm iteration is enhanced, the influence of insufficient space exploration ability on the chameleon swarm algorithm is reduced, and the global exploration ability and convergence speed of the chameleon swarm algorithm are improved. At the same time, combined with the K-means clustering algorithm, an optimal chameleon is finally evolved, and the position of this chameleon is the clustering center of each user cluster. Among them, the improved algorithm is called the chameleon swarm optimization algorithm based on golden sine (GSCSA), and GSCSA is combined with the collaborative filtering algorithm to form an optimized GSCSA-KCF method. This method provides more accurate information for the target user in the user similarity cluster, predicts the product ratings of the user and forms a personalized recommendation list. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The following briefly describes the content expressed in each drawing of the specification of the present invention and the marks in the drawings:

[0036] Figure 1 It is a flowchart of the personalized recommendation method based on clustering and chameleon swarm optimization for the whole-house customization products of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0037] The following further details the specific embodiments of the present invention by describing the optimal embodiments with reference to the drawings.

[0038] The object of the present invention is to provide a personalized recommendation method based on clustering and chameleon swarm optimization for whole-house customization products. This detection method can enhance the global exploration ability of the chameleon swarm algorithm in the early stage of algorithm iteration, reduce the influence of insufficient space exploration ability on the chameleon swarm algorithm, improve the global exploration ability and convergence speed of the chameleon swarm algorithm, further obtain user similarity clusters, recommend users to the target user in its similarity cluster, and give more accurate whole-house customization product recommendation information to the target user.

[0039] As Figure 1 shown, the personalized recommendation method for whole-house customization products based on clustering and chameleon swarm optimization includes the following steps:

[0040] Step 1: Establish a recommendation data set and switch the recommendation data set into a training set and a test set;

[0041] Step 2: Initialize the chameleon population and the number of clusters K;

[0042] Step 3: Use the chameleon swarm algorithm to obtain K optimal similar user clusters;

[0043] Step 4: Use the collaborative filtering algorithm in the test set to rate the products in the similar user clusters of the user to be recommended;

[0044] Step 5: Sort the rated products of the similar user cluster corresponding to the user to be recommended by their equal size, and take the top N products with the largest scores to form a personalized customized product list to recommend products to the user.

[0046] In step 1, the established recommendation dataset is divided into two datasets, namely the training set and the test set; the establishment of the recommendation dataset includes: 1. The recommendation dataset is a data set used to build and evaluate personalized recommendation methods. The dataset includes each user's historical ratings for each item. 2. The establishment of the recommendation dataset is obtained by collecting the rating records of each user on the recommendation platform. 3. The dataset is divided into 70% training set and 30% test set.

[0047] After establishing and dividing the training set and test set in step one, proceed to step two, in which initialization is performed to initialize the chameleon population and the number of clusters K; when evaluating the quality of each chameleon position in each iteration, since each chameleon position represents K cluster centers. Therefore, the position of each chameleon is used as the initial cluster center of the K-means algorithm to cluster all training samples (the content of step three), and then the collaborative filtering algorithm is used in the test set to score the products in the similar user clusters of the recommended users (the content of step four). The above description is represented in the attached figure as the step of "executing the hunting strategy of the chameleon group and evaluating its position".

[0048] In the whole house customization product, the concepts of similar user clusters, products, and cluster centers are as follows. Similar user clusters: When clustering is completed, each cluster consists of similar users. Products: All items involved in the recommended data set. Cluster center: This is a noun, which means the center of each cluster after k-means clustering, which is the average value of all data in the cluster.

[0049] The K-means algorithm is a widely used unsupervised learning algorithm, which is mainly used to divide a data set into several clusters so that the samples within the clusters are as similar as possible and the samples between clusters are as different as possible. Its core idea is to achieve data clustering by iteratively optimizing the cluster centers. The following is the mathematical expression and process of the K-means algorithm.

[0050] Given a set of N samples {x1,x2,…,x N}, the goal of the K-means algorithm is to divide these samples into K clusters C1, C2, ..., C K , so that the distance between each sample and its cluster center is minimized. It is expressed as follows:

[0051]

[0052] Among them, μk It is the cluster center, that is, the clustering center, and is expressed as follows:

[0053]

[0054] In the initial stage of the K-means algorithm, K samples will be randomly selected as the initial centers μ1, μ2, …, μ K For each sample x i , calculate its distances to all clustering centers, and assign it to the cluster center with the minimum distance, which is expressed as follows:

[0055]

[0056] After step two is completed, enter step three, which specifically includes the following steps:

[0057] Step 31: Execute the improved search strategy of the chameleon;

[0058] Step 32: Execute the hunting strategy of the chameleon and evaluate the position;

[0059] Step 33: Determine whether to update the position of the chameleon. If not, retain the current position of the chameleon. If so, update the chameleon population, and then increment the iteration count by one;

[0060] Step 34: Determine whether the maximum number of iterations has been reached; if so, obtain K optimal similar user clusters; if not, return to step 31. The parameters used in the first iteration are all data after initialization.

[0061] Among them, in step 31, the clustering center of the similar user cluster is updated using the improved chameleon search strategy; the improved search strategy of the chameleon includes:

[0062]

[0063] Among them, represents the clustering center position represented by the j-th chameleon at the t-th iteration; Gb t represents the best clustering center position represented by the optimal individual in the chameleon group up to the t-th iteration; up j and low j represent the search space boundaries of the clustering center represented by the j-th chameleon; p1 is a positive number that controls the exploration ability of the algorithm, r and r3 are random numbers within [0, 1], represents the probability that the chameleon detects prey, and the value of sgn(rand - 0.5) is +1 or -1, representing the exploration and exploitation directions; γ, α, and β are three constant values that adjust the exploration and exploitation abilities; r sine1 and r sine2Random numbers between [0, 2π] and [0, π] respectively, representing the moving distance and direction of the individuals in the next generation; x1 = aτ + b(1 - τ) and x2 = a(1 - τ) + bτ, where a and b are -π and π respectively, and τ is the golden ratio number t and T represent the current iteration number and the maximum iteration number respectively.

[0064] In step 32, execute the hunting strategy of the chameleon and evaluate the position; use the chameleon hunting strategy to update the clustering center of the similar user cluster and evaluate the clustering quality of this similar user cluster. All the positions that appear in this scheme are the K clustering centers, because the position of each chameleon individual represents the K clustering centers; when evaluating, the precision of the collaborative filtering recommendation algorithm is adopted.

[0065] The hunting strategy of the chameleon includes:

[0066]

[0067] Among them, represents the speed of the chameleon's tongue when hunting the clustering center at the t-th iteration, a = 2590*(1 - e -log t ) represents the acceleration of the chameleon's tongue popping out. The position of the clustering center represented by the j-th chameleon at the (t + 1)-th iteration after being updated by the hunting strategy.

[0068] Steps 31 and 32 are executed in sequence, first execute 31 and then execute 32, because there is also an eye rotation stage between the search stage and the hunting stage in the entire chameleon swarm algorithm.

[0069] Step 33: Judge whether to update the chameleon position. If not, retain the current chameleon position. If so, update the chameleon population, and then increment the iteration number by one; judge whether to update the position according to the precision of the collaborative filtering recommendation algorithm. If the precision of the collaborative filtering recommendation algorithm in the similar user cluster obtained by the new position is higher, then retain the new position.

[0070] After step three is completed, step four is entered. In step four, in the test set, the collaborative filtering recommendation model is used to score products in the similar user cluster of the user to be recommended; in the test set, the collaborative filtering algorithm is used to score the products that the user to be recommended has not evaluated in its similar user cluster; the positions of the optimal chameleons represent the K clustering centers, and the K-means algorithm is used to form K similar user clusters, and the test set is the user to be recommended; for each user to be recommended, it is necessary to first use K-means to determine which cluster it belongs to, and then perform subsequent operations in this cluster. For the constructed dataset, each row represents the scoring situation of a user for all items. Here, there are products that the user has not evaluated. Therefore, for the user to be recommended, it is necessary to use the scores of other users in its similar user cluster to predict its own scores for the unrated items. The products that have been evaluated do not need to be operated on. At the end of the algorithm, for the user to be recommended, the scores of all its items are sorted, and the top N scores are taken to form a recommendation list. The user to be recommended scores the unrated products in the same cluster, including:

[0071]

[0072] Among them, represents the predicted score of user u for product i, N(u) represents the set of products that user u has historically scored, represents the cosine similarity between user u and user v in the same cluster, I uv represents the set of products that user u and user v in the same cluster have co-scored, I u and I v represent the sets of products that user u and user v in the same cluster have given positive feedback historically, r ui and r vi represent the scores given by user u and user v in the same cluster for product i. Finally, the top N products with the highest scores among the target users are selected to form the personalized whole-house customization product list for user u.

[0073] In step five, the rated products in the similar user cluster corresponding to the user to be recommended are sorted according to the average score, and the top N products with the highest scores are taken to form a personalized customization product list to recommend products to the user, forming a recommendation result, and achieving the purpose of personalized recommendation.

[0074] In this embodiment, the above solution is further elaborated as follows:

[0075] A personalized recommendation method based on clustering and chameleon swarm optimization in whole-house customization products, the recommendation method includes:

[0076] S1. Split the recommendation dataset into a training set and a test set according to a given ratio;

[0077] S2. Initialize the clustering centers and the number of clusters K of similar user clusters; the similar user clusters are to disperse all users into K clusters using the K-means algorithm and the clustering centers.

[0078] S3. Evaluate the clustering quality of each similar user cluster using the precision of the collaborative filtering recommendation algorithm; in S5, it is necessary to compare the current clustering quality with the previous clustering quality. Here, if the clustering quality is not evaluated, the previous clustering quality does not exist in the first iteration. S4. Update the clustering centers of the similar user clusters using the improved chameleon search strategy;

[0079] S5. Update the clustering centers of the similar user clusters using the chameleon hunting strategy and evaluate the clustering quality of the similar user clusters. The chameleon hunting strategy is an iterative process in which poor clustering centers are continuously discarded and excellent clustering centers are retained. At the end of the iteration, an excellent clustering center can enable the k-means algorithm to generate the optimal similar user clusters. In Figure 1 whether to update the chameleon position is to compare the quality of the current clustering center with the quality of the clustering center in the previous iteration. The position of each chameleon represents a set of K clustering centers. S6. When the maximum number of iterations is reached, use the best chameleon individual position in the current population as the K optimal similar user clusters; during the iteration process, the algorithm will continuously discard poor chameleon individual positions and retain excellent chameleon individual positions. There will be an optimal chameleon position after the iteration is completed.

[0080] S7. Use the collaborative filtering algorithm in the test set to score the products that the user to be recommended has not evaluated in its similar user cluster;

[0081] S8. Sort the product ratings of the user to be recommended and select the top N to form a personalized customized product list for personalized whole-house customization.

[0082] Preferably, the initialization of the clustering centers of the similar user clusters in S2 includes:

[0083]

[0084] where, is the initial position of the clustering center represented by the j-th chameleon, up j and low j represent the maximum and minimum boundaries of the accessible search area of the clustering center represented by the j-th chameleon, and r is a random number between 0 and 1.

[0085] Preferably, the precision of the collaborative filtering recommendation algorithm in S3 includes:

[0086]

[0087] Among them, R(u) represents the set of N products recommended to user u, and T(u) represents the set of products that user u likes.

[0088] Preferably, the search strategy for improving the chameleon in S4 includes:

[0089]

[0090] Among them, represents the position of the clustering center represented by the j-th chameleon at the t-th iteration; Gb t represents the best clustering center position represented by the optimal individual in the chameleon group up to the t-th iteration; up j and low j represent the search space boundaries of the clustering center represented by the j-th chameleon; p1 is a positive number that controls the exploration ability of the algorithm, r is a random number within [0, 1], represents the probability that the chameleon detects prey, and the value of sgn(rand - 0.5) is +1 or -1, indicating the direction of exploration and exploitation; γ, α, and β are three constant values that adjust the exploration and exploitation abilities; r sine1 and r sine2 are random numbers between [0, 2π] and [0, π] respectively, representing the moving distance and direction of the next-generation individual; x1 = aτ + b(1 - τ) and x2 = a(1 - τ) + bτ, where a and b are -π and π respectively, and τ is the golden ratio t and T represent the current iteration number and the maximum iteration number respectively.

[0091] Preferably, the hunting strategy of the chameleon in S5 includes:

[0092]

[0093] Among them, represents the speed of the chameleon's tongue when hunting the clustering center, and a = 2590*(1 - e -log t ) represents the acceleration of the chameleon's tongue popping out.

[0094] Preferably, in S7, the unrated products are scored by the users to be recommended in the same cluster, including:

[0095]

[0096] Among them, represents the predicted score of user u for product i, and N(u) represents the set of products with the historical scores of user u, represents the cosine similarity between user u and user v in the same cluster, I uvDenote the set of products jointly rated by user u and user v in the same cluster as I u and I v Denote the set of products that user u and user v in the same cluster have given positive feedback on historically as r ui and r vi Denote the ratings given by user u and user v in the same cluster to product i. Finally, select the top N products rated by the target user to form the personalized whole-house customization product list for user u.

[0097] According to the above technical solution, the present invention introduces the golden sine strategy in the prey search stage of the chameleon swarm algorithm, improves the position search equation of the chameleon swarm in this stage, enhances the global exploration ability of the chameleon swarm algorithm in the early stage of algorithm iteration, reduces the impact of insufficient space exploration ability on the chameleon swarm algorithm, and improves the global exploration ability and convergence speed of the chameleon swarm algorithm. At the same time, combined with the K-means clustering algorithm, an optimal chameleon is finally evolved, and the position of this chameleon is the clustering center of each user cluster. Among them, the improved algorithm is called the golden-sine-based chameleon swarm optimization algorithm (GSCSA), and GSCSA is combined with the collaborative filtering algorithm to form the optimized GSCSA-KCF method. This method provides more accurate information for the target user in the user similarity cluster, predicts the user's product ratings and forms a personalized recommendation list.

[0098] In this field, the swarm intelligence optimization method, as a method for solving optimization problems, can imitate the behaviors of animals such as hunting and reproduction to discover the optimal solution in space, thereby improving the recommendation efficiency and accuracy of the recommendation algorithm.

[0099] According to the above technical solution, the present invention introduces the golden sine strategy in the prey search stage of the chameleon swarm algorithm, improves the position search equation of the chameleon swarm in this stage, enhances the global exploration ability of the chameleon swarm algorithm in the early stage of algorithm iteration, reduces the impact of insufficient space exploration ability on the chameleon swarm algorithm, and improves the global exploration ability and convergence speed of the chameleon swarm algorithm. At the same time, combined with the K-means clustering algorithm, an optimal chameleon is finally evolved, and the position of this chameleon is the clustering center of each user cluster. Among them, the improved algorithm is called the golden-sine-based chameleon swarm optimization algorithm (GSCSA), and GSCSA is combined with the collaborative filtering algorithm to form the optimized GSCSA-KCF method. This method provides more accurate information for the target user in the user similarity cluster, predicts the user's product ratings and forms a personalized whole-house customization recommendation list.

[0100] The above CSA algorithm is a new type of swarm intelligence optimization algorithm proposed by Braik in 2021. This algorithm mainly simulates the foraging and hunting behaviors of chameleons in desert and jungle environments. The specific description of this algorithm is as follows:

[0101] 1. Searching for prey stage: When searching for prey, the position of a chameleon individual can be updated using the following formula:

[0102]

[0103] Where, represents the position of the cluster center represented by the j-th chameleon at the t-th iteration; Gb t represents the best cluster center position represented by the optimal individual in the chameleon group up to the t-th iteration; up j and low j represent the boundaries of the search space for the cluster center represented by the j-th chameleon; p1 is a positive number that controls the exploration ability of the algorithm, r is a random number within [0, 1], represents the probability that a chameleon detects prey, and the value of sgn(rand - 0.5) is +1 or -1, indicating the direction of exploration and exploitation; γ, α, and β are three constant values that regulate the exploration and exploitation abilities; r sine1 and r sine2 are random numbers between [0, 2π] and [0, π] respectively, representing the moving distance and direction of the next-generation individual; x1 = aτ + b(1 - τ) and x2 = a(1 - τ) + bτ, where a and b are -π and π respectively, and τ is the golden ratio t and T represent the current iteration number and the maximum iteration number respectively.

[0104] 2. Eye rotation stage: Chameleons can sense the position of prey by rotating their eyes, which enables them to observe prey within a 360-degree range. Once a chameleon discovers prey, it will move in the direction of the prey. In this case, the formula for updating the chameleon's position is as follows:

[0105]

[0106] Where, represents the average position of the chameleon population at the t-th iteration; represents the rotation center of the j-th chameleon at the t-th iteration, which can be expressed by the following formula:

[0107]

[0108] Where, represents the central coordinates of the j-th chameleon at the t-th iteration, represents the chameleon rotation matrix. Where, θ = r * sgn(rand - 0.5) * 180° represents the rotation angle of the chameleon's eyes, and represent a set of orthogonal vectors, A x and Ay They respectively represent the rotation matrices on the x-axis and y-axis, and the expressions are as follows:

[0109]

[0110] Among them, represents the rotation angle around the x-axis, represents the rotation angle around the y-axis.

[0111] 3. During the stage of hunting prey, when chameleons approach their prey, they will attack it with their tongues. The chameleon closest to the prey is considered the optimal chameleon and is also considered the optimal solution in the solution space. During the process of hunting for the global optimal clustering center, the tongue of the chameleon extends to twice its original length. The speed at which the tongue is ejected towards the prey can be expressed as follows:

[0112]

[0113] Among them, is the tongue ejection speed of the j-th chameleon at the t-th iteration, c1 and c2 are two positive numbers, represents the inertia weight, which decreases as the number of iterations increases, and ρ is a positive number used to affect the exploration ability. When the chameleon extends its tongue towards the direction of the clustering center, its own position will also change accordingly. The formula for this position change is as follows:

[0114]

[0115] Among them, a = 2590*(1 - e -log t ) represents the acceleration of the chameleon's tongue ejection.

[0116] In the present invention, the user-based collaborative filtering algorithm (User-CF) is a commonly used recommendation algorithm. Its core idea is that similar users have similar ratings and preferences for products in history. Therefore, other users similar to the target user can be found, and the products liked by these similar users can be recommended. At this time, the following cosine similarity formula is used to calculate the similarity between users:

[0117]

[0118] Among them, Sim(u, v) represents the cosine similarity between users u and v, I uv represents the set of products jointly rated by users u and v, I u and I v represent the sets of products that users u and v have given positive feedback on in history, r ui and r vi represent the ratings given by users u and v to the customized product i.

[0119] Subsequently, predict the scores of products that have not been rated among the users to be recommended:

[0120]

[0121] Among them, represents the predicted score of user u for product i, and N(u) represents the set of customized products for which user u has historically rated products. Finally, select the top N products with ratings among the target users to form the personalized list of user u.

[0122] It can be seen from this that the GSCSA method has the ability to enhance the global exploration ability of the chameleon swarm algorithm in the early stage of algorithm iteration, reduce the impact of insufficient space exploration ability on the chameleon swarm algorithm, improve the global exploration ability and convergence speed of the chameleon swarm algorithm. Therefore, the above-mentioned personalized recommendation method based on clustering and chameleon swarm optimization in the whole-house customization products is proposed, and:

[0123] The initialization of the clustering center of similar user clusters in S2 includes:

[0124]

[0125] Among them, is the initial position of the clustering center represented by the j-th chameleon, up j and low j represent the maximum and minimum boundaries of the accessible search area of the clustering center represented by the j-th chameleon, and r is a random number between 0 and 1.

[0126] In S3, the precision of the collaborative filtering recommendation algorithm includes:

[0127]

[0128] Among them, R(u) represents the set of N products recommended to user u, and T(u) represents the set of products that user u likes.

[0129] In S4, the improved search strategy of the chameleon includes:

[0130]

[0131] Among them, represents the position of the clustering center represented by the j-th chameleon at the t-th iteration; Gb t represents the position of the best clustering center represented by the optimal individual in the chameleon swarm up to the t-th iteration; up j and low j represent the search space boundaries of the clustering center represented by the j-th chameleon; p1 is a positive number that controls the exploration ability of the algorithm, and r is a random number in [0,1]. Represents the probability that the chameleon detects prey, and the value of sgn(rand - 0.5) is +1 or -1, indicating the direction of exploration and exploitation; γ, α, and β are three constant values for adjusting the exploration and exploitation capabilities; r sine1 and r sine2 are random numbers between [0, 2π] and [0, π] respectively, representing the movement distance and direction of the next-generation individuals; x1 = aτ + b(1 - τ) and x2 = a(1 - τ) + bτ, where a and b are -π and π respectively, and τ is the golden ratio t and T represent the current iteration number and the maximum iteration number respectively.

[0132] In S5, the hunting strategy of the chameleon in S5 includes:

[0133]

[0134] Among them, represents the speed of the chameleon's tongue when hunting the clustering center, and a = 2590*(1 - e -log t ) represents the acceleration of the chameleon's tongue popping out.

[0135] In S7, the operation of the user to be recommended to score the unrated products in the same cluster includes:

[0136]

[0137] Among them, represents the predicted score of user u for product i, and N(u) represents the set of products that user u has historically scored for products, represents the cosine similarity between user u and user v in the same cluster, I uv represents the set of products that user u and user v in the same cluster have co - rated, I u and I v represent the sets of products that user u and user v in the same cluster have given positive feedback on historically, r ui and r vi represent the scores given by user u and user v in the same cluster for the customized product i. Finally, the top N products with the highest scores among the target users are selected to form the personalized list of user u.

[0138] The preferred embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited to the specific details in the above - mentioned embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solutions of the present invention, and these simple modifications all fall within the protection scope of the present invention.

[0139] In addition, it should be noted that, in the above specific embodiments, the various specific technical features described can be combined in any appropriate manner without contradiction. To avoid unnecessary repetition, the present invention will not separately describe various possible combinations.

[0140] Obviously, the specific implementation of the present invention is not limited by the above methods. As long as various non-substantive improvements are made by adopting the method concept and technical solution of the present invention, they are all within the protection scope of the present invention.

Claims

1. A personalized recommendation method for whole-house customization products based on clustering and chameleon swarm optimization, characterized in that: It includes the following steps: S1. Establish a recommendation dataset and switch the recommendation dataset into a training set and a test set; S2. Initialize the chameleon population and the number of clusters K; S3. Use the chameleon swarm algorithm to obtain K optimal similar user clusters; S4. In the test set, use the collaborative filtering algorithm to rate products in the similar user clusters of the user to be recommended; S5. Sort the rated products of the similar user clusters corresponding to the user to be recommended according to the average score, and select the top N with higher scores to form a personalized customized product list for product recommendation to the user.

2. The personalized recommendation method for the whole-house customized products by clustering and chameleon swarm optimization according to claim 1, characterized in that: Step S2 includes: Initialize the cluster centers of the similar user clusters and the number of clusters K, and use the precision of the collaborative filtering recommendation algorithm to evaluate the clustering quality of each similar user cluster.

3. The personalized recommendation method for the whole-house customized products by clustering and chameleon swarm optimization according to claim 2, characterized in that: The initialization of the cluster centers in step S2 includes: Initialize using the following formula: Among them, is the initial position of the clustering center represented by the j-th chameleon, up j and low j represent the maximum and minimum boundaries of the accessible search area of the clustering center represented by the j-th chameleon, and r is a random number between 0 and 1.

4. The personalized recommendation method for whole-house customization products based on clustering and chameleon swarm optimization according to claim 2, wherein: What is included in step S3 is: S31. Update the cluster centers of the similar user clusters using the improved chameleon search strategy; S32. Update the cluster centers of the similar user clusters using the chameleon hunting strategy and evaluate the clustering quality of the similar user clusters; S33. Determine whether the iteration number threshold is reached. If not, feedback to S31; if so, use the position of the best chameleon individual in the current population as the K optimal similar user clusters.

5. The personalized recommendation method for whole-house customized products based on clustering and chameleon swarm optimization according to claim 2, wherein: The precision of the collaborative filtering recommendation algorithm includes: Among them, R(u) represents the set of N products recommended to user u, and T(u) represents the set of products liked by user u.

6. The personalized recommendation method for whole-house customized products based on clustering and chameleon swarm optimization according to claim 4, wherein: The improved chameleon search strategy includes: Among them, represents the position of the clustering center represented by the j-th chameleon at the t-th iteration; Gb t represents the best clustering center position represented by the optimal individual in the chameleon group up to the t-th iteration; up j and low j represent the search space boundaries of the clustering center represented by the j-th chameleon; p1 is a positive number that controls the exploration ability of the algorithm, r is a random number within [0,1], represents the probability that the chameleon detects prey, the value of sgn(rand - 0.5) is +1 or -1, indicating the direction of exploration and exploitation; γ, α, and β are three constant values that adjust the exploration and exploitation abilities; r sine1 and r sine2 are random numbers between [0, 2π] and [0, π] respectively, representing the moving distance and direction of the next-generation individual; x1 = aτ + b(1 - τ) and x2 = a(1 - τ) + bτ, where a and b are -π and π respectively, and τ is the golden ratio t and T represent the current iteration number and the maximum iteration number respectively.

7. The personalized recommendation method for whole-house customized products based on clustering and chameleon swarm optimization according to claim 4, wherein: The chameleon hunting strategy includes: Among them, represents the speed of the chameleon's tongue when hunting the clustering center, and a = 2590*(1 - e -logt ) represents the acceleration of the chameleon's tongue when it pops out.

8. The personalized recommendation method for whole-house customized products based on clustering and chameleon swarm optimization according to any one of claims 1-7, wherein: The user to be recommended scores the unrated products in the same cluster including: Among them, represents the predicted score of user u for product i, and N(u) represents the set of products that user u has historically scored for products, represents the cosine similarity between user u and user v in the same cluster, and I uv represents the set of products that user u and user v in the same cluster have jointly rated, and I u and I v represent the set of products for which user u and user v in the same cluster have given positive feedback historically, r ui and r vi represent the ratings given by user u and user v in the same cluster for product i. Finally, the top N products with the highest ratings among the target users are selected to form the personalized whole-house customization product list for user u.

Images