Recommended methods for psychological assessment scales

Abstract

This invention discloses a method for recommending psychological assessment scales, comprising the following steps: Step 1, acquiring historical psychological assessment data and preprocessing the data; Step 2, constructing a user-scale assessment score matrix based on the preprocessed data from Step 1; Step 3, constructing user clusters using the K-means algorithm based on the user-scale assessment score matrix from Step 2; Step 4, when recommending scales to target users, calculating the similarity between the target user and the centroids of each cluster obtained in Step 3, selecting the user cluster containing the centroid with the highest similarity as the cluster to which the target user belongs, and generating a set of recommended scales based on the partitioned clusters. This invention can automatically recommend suitable psychological assessment scales to target users, protect users' personal information, improve the accuracy, diversity, and novelty of recommendations, and simultaneously use particle swarm optimization K-means algorithm to reduce the problem size, improve recommendation speed, and save computational resources.

Description

Technical Field

[0001] This invention belongs to the technical field of psychological testing, specifically relating to a method for recommending psychological testing scales. Background Technology

[0002] Psychological Assessment Systems: With increasing societal focus on mental health issues, organizations such as businesses and schools have an urgent need to understand the mental health status of their employees. Failure to promptly identify mental health problems can lead to risks to personal safety, health, and property. Traditional psychological assessments rely on professional psychoanalysts who communicate with clients, analyzing their mental health based on the analyst's experience, the client's performance, and the assessment questions. Online psychological assessment systems, on the other hand, require users to select from a large number of assessment scales. Recommending suitable assessment scales to target users can greatly improve the efficiency of psychological assessments, saving time and manpower, and enabling timely acquisition of mental health information.

[0003] Psychological assessment scales are sets of questions summarized by psychology experts, based on comprehensive analysis of diagnostic data, and designed to address specific psychological issues. They serve as a measurement tool for assessing an individual's mental health. Because psychological assessment scales are generally based on statistical analysis, they often have universal applicability. Psychological assessment scales are divided into single-dimensional scales and multi-dimensional scales, which assess a user's mental health based on a single score and multiple scores, respectively. For psychological assessment scales used to warn of potential mental health problems, a higher score indicates a higher likelihood that the user has that particular mental health issue.

[0004] Historical records of psychological assessments: Online psychological assessment systems store the historical psychological assessment records of each user, including assessment duration, assessment scale, assessment results, and other items. This structured data forms the data basis for recommending psychological assessment scales.

[0005] Recommendations for Psychological Assessment Scales: A wide variety of psychological assessment scales exist across different fields of psychology. Traditional methods heavily rely on psychoanalysts, and when using online psychological assessment systems, users often struggle to select the appropriate scale from a vast array. If the effectiveness of the assessment scales cannot be guaranteed, the accuracy of assessing a user's mental health becomes difficult, leading to unnecessary waste of resources. Therefore, recommending suitable psychological assessment scales for individual users is essential.

[0006] Current methods for recommending psychological assessment scales mostly rely on demographic information such as user occupation, age, and gender to suggest a batch of potentially applicable scales. These methods fail to consider the influence of subjective factors on mental health, and the single data source leads to biased recommendations, making accuracy difficult to guarantee. Furthermore, these methods require access to user demographic information, which compromises user privacy. Therefore, there is an urgent need to develop new methods for recommending psychological assessment scales to address these issues. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of existing technologies by providing a method for recommending psychological assessment scales. This method proposes a collaborative filtering-based approach for recommending psychological assessment scales and uses a particle swarm optimization K-means algorithm for clustering to reduce the problem size, thereby improving the data sparsity problem in collaborative filtering and ultimately completing the recommendation of psychological assessment scales.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0009] A recommended method for a psychological assessment scale includes the following steps:

[0010] Step 1: Obtain historical psychological assessment data and preprocess the data;

[0011] Step 2: Construct a user-scale score matrix based on the preprocessed data from Step 1;

[0012] Step 3: Based on the user-scale score matrix from Step 2, construct user clusters using the K-means algorithm;

[0013] Step 4: When recommending scales to target users, calculate the similarity between the target user and each cluster center point obtained in Step 3, take the user cluster containing the center point with the highest similarity as the cluster to which the target user belongs, and generate a set of recommendation scales based on the divided clusters.

[0014] Furthermore, the data obtained in step 1 is as follows:

[0015] This system uses Python to read and clean historical records of psychological assessments stored in a database. Each historical record represents a user's score on a psychological assessment scale.

[0016] R i ={U t ,S j ,P i}

[0017] Among them, R i U represents the i-th historical evaluation record.t S represents the user ID. j Indicates scale ID, P i This indicates the score.

[0018] Furthermore, the preprocessing method is as follows:

[0019] For a multidimensional scale with multiple test scores, each dimension is split into a single-dimensional scale and treated as a separate historical record; the scale test scores in the historical record are normalized and converted into a percentage system in equal proportions.

[0020] Furthermore, step 3 specifically includes:

[0021] Step 3.1: Randomly select K users as initial center points and calculate user U. t Similarity with each center point, and user U t Assign them to the cluster containing the centroid with the highest similarity;

[0022] Step 3.2: Continue iterative calculation. Before reaching the maximum number of iterations M, repeat the steps of recalculating the center point of each cluster and assigning users to the cluster containing the center point with the highest similarity.

[0023] Step 3.3: When the maximum number of iterations M is reached, the iteration ends, and K clusters are generated and output, including the centroids and user sets;

[0024] Step 3.4: Use the particle swarm optimization algorithm to iteratively calculate the K clusters divided in Step 3.3, and output a set of cluster center points as the initial center points for the K-means algorithm.

[0025] Furthermore, in step 3.1, cosine similarity is used to calculate the similarity between the two users:

[0026]

[0027] Among them, U i U j These are two users whose similarity needs to be calculated. S is the scale, and Q is the user's score. i The psychological assessment scales and U tested j The intersection of the psychological assessment scales used, P U,S It is the score of user U on psychological assessment scale S, corresponding to an item in the user-scale score matrix;

[0028] The range of cosine similarity is [-1, 1]. The closer the cosine similarity value is to 1, the higher the similarity between the two users.

[0029] Furthermore, the method for iteratively calculating and dividing the new family in step 3.2 is as follows:

[0030] Calculate the centroid of each cluster, and for each cluster, calculate the mean similarity between each user and other users:

[0031]

[0032] Where W is the mean similarity between user U and other users in the cluster, and m is the number of users in the cluster;

[0033] Using the user with the smallest mean as the new center point, the remaining users are divided according to step 3.1 to form a new cluster.

[0034] Furthermore, the method for calculating the initial centroids of the K-means algorithm using the particle swarm optimization (POS) iterative method in step 3.4 is as follows:

[0035] The fitness function of the Particle Swarm Optimization (PSO) algorithm is defined as finding K cluster centers such that the fitness of all users to their respective cluster centers is maximized.

[0036]

[0037] Among them, D k It is a dataset. The original data is divided into K clusters to obtain a new dataset, μ = {μ1, μ2, ..., μ...} K} represents the cluster centers, and n is the number of data entries in the k-th cluster. It is the i-th data in the k-th cluster;

[0038] PSO initializes a set of random cluster centers and treats them as random particles. It calculates the fitness of the initial random solution and initializes the individual optimal position, individual optimal fitness, population optimal position, and population optimal fitness.

[0039] The optimal solution is found through iteration. In each iteration, the particle updates its velocity and position by tracking.

[0040] Check and update the individual optimal position, individual optimal fitness, group optimal position, and group optimal fitness until the maximum number of iterations is reached. Output a set of cluster centers as the initial centers for the K-means algorithm.

[0041] Furthermore, step 4 also includes the following sub-steps:

[0042] Step 4.1: In this cluster, construct a user-scale score matrix and calculate the similarity between the target user and other users in the cluster. Select the L users with the highest similarity as the target user's nearest neighbors.

[0043] Step 4.2: Predict the scores of target users on various psychological assessment scales;

[0044] Step 4.3: After obtaining the target user's predicted scores for each psychological assessment scale, merge the multiple single-dimensional scales that were split from a multidimensional scale in Step 1. Take the highest predicted score of the multiple single-dimensional scales as the predicted score of the multidimensional scale. Sort the psychological assessment scales in descending order according to the target user's predicted scores, and take the top N scales as the target user's recommended scales for output.

[0045] Furthermore, the formula for calculating the score of the predictive scale is as follows:

[0046]

[0047] Among them, F U,S It is the predicted score of user U on scale S. sim(U, U) is the average score of users on the scale, C is the set of nearest neighbor users of the target user U, and sim(U, U) is the average score of users on the scale. j ) represents the target user U and the nearest neighbor user U. j cosine similarity, P U,S It is the score of user U on scale S, that is, an item in the user-scale score matrix.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention proposes a method for recommending psychological assessment scales based on collaborative filtering, and uses the particle swarm optimization K-means algorithm for clustering to reduce the problem size, thereby improving the problem of data sparsity in collaborative filtering, and finally completing the recommendation of psychological assessment scales;

[0049] This invention can automatically recommend suitable psychological assessment scales to target users, using user behavioral data without relying on user personal information, thus protecting user personal information, improving the accuracy, diversity, and novelty of recommendations, and increasing the likelihood of discovering users' mental health problems. At the same time, it uses the particle swarm optimization K-means algorithm to reduce the problem size, improve recommendation speed, and save computing resources. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating the implementation of the recommendation method for the psychological assessment scale in this invention. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0052] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0053] The present invention will be further described below with reference to specific embodiments, but these are not intended to limit the scope of the invention.

[0054] like Figure 1 As shown, this invention provides a method for recommending psychological assessment scales, comprising the following steps:

[0055] Step 1: Obtain historical psychological assessment data and preprocess the data;

[0056] In this embodiment, Python is used to read and clean the psychological assessment history records stored in the database. Each history record represents a user's score on a psychological assessment scale.

[0057] R i ={U t ,S j ,P i}

[0058] Among them, R i U represents the i-th historical evaluation record. t S represents the user ID. j Indicates scale ID, P i This indicates the score.

[0059] For a multidimensional scale with multiple test scores, each dimension is split into a single-dimensional scale as a separate historical record. The scale test scores in the historical record are then normalized and converted into a percentage system.

[0060] Step 2: Construct a user-scale score matrix based on the preprocessed data from Step 1;

[0061] Based on the data preprocessed in step 1, a user-scale rating matrix is ​​constructed. The constructed user-scale rating matrix is ​​shown in Table 1.

[0062] Table 1.1 User-scale rating matrix

[0063]

[0064]

[0065] In Table 1, U represents users, S represents scales, and P represents... U,S Let U be the score of user U on scale S. In psychological assessment systems, users often have multiple historical assessment records for a scale. The average score of these historical assessment records is used to calculate a matrix.

[0066] Step 3: Based on the user-scale score matrix from Step 2, construct user clusters using the K-means algorithm;

[0067] Step 3.1: Construct the initial clusters;

[0068] Randomly select K users as initial center points and calculate user U. t The similarity between two users is calculated using cosine similarity, based on the similarity at each center point:

[0069]

[0070] Among them, U i U j These are two users whose similarity needs to be calculated. S is the scale, and Q is the user's score. i The psychological assessment scales and U tested j The intersection of the psychological assessment scales used, P U,S It is the score of user U on psychological assessment scale S, corresponding to an item in the user-scale score matrix;

[0071] User U t Assign them to the cluster containing the centroid with the highest similarity;

[0072] Step 3.2: Continue iterating. Before reaching the maximum number of iterations M, repeat the steps of recalculating the center point of each cluster and assigning users to the cluster containing the center point with the highest similarity.

[0073] Calculate the centroid of each cluster, and for each cluster, calculate the mean similarity between each user and other users:

[0074]

[0075] Where W is the mean similarity between user U and other users in the cluster, and m is the number of users in the cluster;

[0076] Using the user with the smallest mean as the new center point, the remaining users are divided according to step 3.1 to form a new cluster;

[0077] Step 3.3: Generate the final clusters. The iteration ends when the maximum number of iterations M is reached, and K clusters are output, including the centroids and user sets.

[0078] Step 3.4: The particle swarm optimization algorithm is used to iteratively calculate the K clusters divided in step 3.3, and output a set of cluster center points as the initial center points for the K-means algorithm.

[0079] Particle Swarm Optimization (PSO) simulates birds in a flock by designing massless particles that have only two attributes: velocity v. i and current position x i Speed ​​represents the rate of movement, and position represents the direction of movement. Each particle has a fitness value determined by the objective function and knows its best position so far, pbest, and the best position so far found by the swarm, gbest.

[0080] The fitness function of PSO is defined as finding K cluster centers such that all users have the highest fitness to their respective cluster centers:

[0081]

[0082] Among them, D k It is a dataset. The original data is divided into K clusters to obtain a new dataset, μ = {μ1, μ2, ..., μ...} K} represents the cluster centers, and n is the number of data entries in the k-th cluster. It is the i-th data in the k-th cluster.

[0083] PSO initializes a set of random cluster centers, treating them as random particles, calculates the fitness of the initial random solution, and initializes the individual optimal position, individual optimal fitness, population optimal position, and population optimal fitness.

[0084] The optimal solution is found through iteration. In each iteration, the particle tracks (p...) best ,g best Update your speed and position using the following formula:

[0085] v i (t+1)=v i (t)+c1×rand()×(pbest i (t)-x i (t))+c2×rand()

[0086] ×(gbest(t)-x i (t))

[0087] x i (t+1)=x i (t)+v i (t+1)

[0088] Where i = 1, 2, ..., N, N is the total number of particles in the swarm, rand() is a random number between (0, 1), c1 and c2 are learning factors, set c1 = c2 = 2, v i The maximum value is V max (Set as the range width of the particle), if v i Greater than V max Then v i =V max .

[0089] An inertia weighting factor is introduced to balance the effects of local and global searches and optimize the convergence speed. Therefore, the formulas for particle update speed and position are as follows:

[0090] v i (t+1)=w×v i (t)+c1×rand()×(pbest i (t)-x i (t))+c2×rand()

[0091] ×(gbest(t)-x i (t))

[0092] x i (t+1)=x i (t)+v i (t+1)

[0093] Where w is the inertia factor, its value is non-negative, and the value of w is set using a linearly decreasing weighting strategy:

[0094]

[0095] Among them, G max It is the maximum number of iterations, w int This initializes the weights to 0.9, w. end It is the weight when the iteration reaches its maximum, and it is set to 0.4.

[0096] Check and update the individual optimal position, individual optimal fitness, group optimal position, and group optimal fitness until the maximum number of iterations is reached. Output a set of cluster centers as the initial centers for the K-means algorithm.

[0097] Step 4: When recommending rating scales to the target user, calculate the similarity between the target user and each cluster centroid obtained in Step 3, take the user cluster containing the centroid with the highest similarity as the cluster to which the target user belongs, and generate a set of recommendation rating scales based on the divided clusters; in this embodiment, this step specifically includes:

[0098] Step 4.1: Calculate user similarity and select nearest neighbor users; In this cluster, construct a user-scale score matrix, calculate the similarity between the target user and other users in the cluster based on cosine similarity, and select the L users with the highest similarity as the target user's nearest neighbor users;

[0099] Step 4.2: Predict scale scores; predict the target user's scores on various psychological assessment scales. The formula for calculating the predicted scale scores is as follows:

[0100]

[0101] In the formula, F U,S It is the predicted score of user U on scale S. sim(U, U) is the average score of users on the scale, C is the set of nearest neighbor users of the target user U, and sim(U, U) is the average score of users on the scale. j ) represents the target user U and the nearest neighbor user U. j cosine similarity, P U,S It is the score of user U on scale S, that is, an item in the user-scale score matrix;

[0102] Step 4.3: Recommend scales. After obtaining the target user's predicted scores for each psychological assessment scale, merge the multiple unidimensional scales derived from the multidimensional scale in Step 1. Take the highest predicted score from each unidimensional scale as the predicted score for that multidimensional scale. Sort the psychological assessment scales in descending order of the target user's predicted scores, and select the top N scales as the recommended scales for the target user, then output them.

[0103] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made based on the content of this specification should be included within the protection scope of the present invention.

Claims

1. A method for recommending a psychological assessment scale, characterized in that, Includes the following steps: Step 1: Obtain historical psychological assessment data and preprocess the data; Step 2: Construct a user-scale score matrix based on the preprocessed data from Step 1; Step 3: Based on the user-scale score matrix from Step 2, construct user clusters using the K-means algorithm; Step 4: When recommending scales to target users, calculate the similarity between the target user and each cluster center point obtained in Step 3, take the user cluster where the center point with the highest similarity is located as the cluster to which the target user belongs, and generate a set of recommendation scales based on the divided clusters. Step 3 specifically includes: Step 3.1, Random selection Using each user as the initial center point, calculate the user... Similarity with each center point, and user Assign them to the cluster containing the centroid with the highest similarity; Step 3.2: Continue iterative calculation until the maximum number of iterations is reached. Previously, the steps of recalculating the center point of each cluster and assigning users to the cluster containing the center point with the highest similarity were repeated. Step 3.3: When the maximum number of iterations M is reached, the iteration ends, and K clusters are generated and output, including the centroids and user sets; Step 3.4: Use the particle swarm optimization algorithm to iteratively calculate the K clusters divided in Step 3.3, and output a set of cluster center points as the initial center points for the K-means algorithm; The method for calculating the initial centroids of the K-means algorithm using the particle swarm optimization (POS) iterative method in step 3.4 is as follows: The fitness function of the Particle Swarm Optimization (PSO) algorithm is defined as finding K cluster centers such that the fitness of all users to their respective cluster centers is maximized. in, It's a dataset; the original data is divided into K clusters to obtain a new dataset. The cluster centers are the data points in the k-th cluster. It is the i-th data in the k-th cluster; PSO initializes a set of random cluster centers and treats them as random particles. It calculates the fitness of the initial random solution and initializes the individual optimal position, individual optimal fitness, population optimal position, and population optimal fitness. The optimal solution is found through iteration. In each iteration, the particle updates its velocity and position by tracking. Check and update the individual optimal position, individual optimal fitness, group optimal position, and group optimal fitness until the maximum number of iterations is reached. Output a set of cluster centers as the initial centers for the K-means algorithm.

2. The method for recommending the psychological assessment scale according to claim 1, characterized in that, The data obtained in step 1 is: This system uses Python to read and clean historical records of psychological assessments stored in a database. Each historical record represents a user's score on a psychological assessment scale. in, This represents the i-th historical evaluation record. Represents user ID, Indicates the scale ID. This indicates the score.

3. The method for recommending the psychological assessment scale according to claim 2, characterized in that, The preprocessing method is as follows: For a multidimensional scale with multiple test scores, each dimension is split into a single-dimensional scale and treated as a separate historical record; the scale test scores in the historical record are normalized and converted into a percentage system in equal proportions.

4. The method for recommending the psychological assessment scale according to claim 1, characterized in that, Step 3.1 uses cosine similarity to calculate the similarity between the two users: in, , These are two users whose similarity needs to be calculated. It's a scale. yes Psychological assessment scales and The intersection of previously administered psychological assessment scales. It is the score of user U on psychological assessment scale S, corresponding to an item in the user-scale score matrix; The range of cosine similarity is The closer the cosine similarity value is to 1, the higher the similarity between the two users.

5. The method for recommending the psychological assessment scale according to claim 1, characterized in that, The method for iteratively calculating and dividing the new family in step 3.2 is as follows: Calculate the centroid of each cluster, and for each cluster, calculate the mean similarity between each user and other users: in, For users The mean similarity with other users in the cluster. The number of users in the cluster; Using the user with the smallest mean as the new center point, the remaining users are divided according to step 3.1 to form a new cluster.

6. The method for recommending the psychological assessment scale according to claim 1, characterized in that, Step 4 also includes the following sub-steps: Step 4.1: In this cluster, construct a user-scale score matrix and calculate the similarity between the target user and other users in the cluster. Select the L users with the highest similarity as the target user's nearest neighbors. Step 4.2: Predict the scores of target users on various psychological assessment scales; Step 4.3: After obtaining the target user's predicted scores for each psychological assessment scale, take the highest predicted score of each single-dimensional scale as the predicted score of that multidimensional scale. Sort each psychological assessment scale in descending order according to the target user's predicted scores, and take the top N scales as the target user's recommended scale for output.

7. The method for recommending the psychological assessment scale according to claim 6, characterized in that, The formula for calculating the score of the predictive scale is as follows: in, It is the predicted score of user U on scale S. C is the average score of users on the scale, and C is the set of nearest neighbor users of the target user U. It is the target user U and the nearest neighbor users cosine similarity, It is the score of user U on scale S, that is, an item in the user-scale score matrix.

Images