Behavioral risk tendency prediction method and system based on subjective and objective information fusion representation

Abstract

The application belongs to the cross field of deep learning and prediction, and provides a behavior risk tendency prediction method and system based on subjective and objective information fusion representation. According to the subjective and objective fusion information of the to-be-evaluated object and the information reconstruction network, corresponding data coding is obtained. Based on the data set composed of the data coding, the data set is divided into multiple subsets. Based on each data subset, a classification algorithm stored in an algorithm warehouse is used, the accuracy and micro F1 value of each classification algorithm are determined, the classification algorithm suitable for all sample data is determined, the algorithm label of the sample data is obtained, and a database for storing the subjective and objective information fusion representation with the algorithm label is constructed. Based on the database data and test data, the nearest neighbor node algorithm is used for data comparison, a suitable classification algorithm is selected and applied to the test data, and a behavior prediction result is obtained. It is suitable for different to-be-evaluated objects, different environments and different subjective and objective information, so that the generalization ability of the algorithm is improved.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of deep learning and evaluation science, and in particular relates to a method and system for predicting behavioral risk tendencies based on the fusion of subjective and objective information. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Behavioral prediction methods can be divided into two categories: medical clinical assessment and statistical assessment. Medical clinical assessment requires clinical observation by professional psychologists, and the results largely depend on the assessor's theoretical knowledge and practical observation. This method is less efficient and faces challenges in its widespread adoption. Statistical assessment, on the other hand, is quantifiable and has clearly defined evaluation criteria. This method comprehensively considers multi-dimensional variables associated with behavioral risk propensity levels, resulting in more accurate assessments.

[0004] Existing methods for predicting behavioral risk tendencies have the following shortcomings:

[0005] Feature selection is a crucial step in risk propensity prediction. It involves extracting meaningful features from the raw data for the model to learn and predict. One of the challenges of feature selection is identifying features relevant to the target behavior in the raw data while excluding irrelevant or redundant features. Traditional methods often struggle to address this issue; selecting too many or irrelevant features can lead to the curse of dimensionality, making the model more complex, increasing computational costs, and potentially causing overfitting and reducing the model's generalization ability.

[0006] Furthermore, it's difficult for any single algorithm to be applicable to all situations. Therefore, the specific problem and data characteristics must be considered when selecting a model. Incorrect model selection can lead to performance degradation or inaccurate predictions. However, most existing methods for predicting behavioral risk tendencies employ fixed algorithms, resulting in poor applicability to different behavioral prediction targets and environments. Summary of the Invention

[0007] To address at least one of the technical problems existing in the background art, the present invention provides a method and system for predicting behavioral risk tendencies based on the fusion of subjective and objective information. It achieves comprehensive utilization of subjective and objective information through deep learning, reduces the influence of human factors, and aims to conduct a comprehensive and objective assessment of behavioral risk tendencies, saving human and material resources and improving assessment efficiency.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] The first aspect of the present invention provides a method for predicting behavioral tendencies based on the fusion of subjective and objective information, comprising the following steps:

[0010] Obtain subjective and objective information of the object to be predicted, and process it to obtain the integrated subjective and objective information of the object to be predicted.

[0011] By combining subjective and objective information with trained data, the network is reconstructed to obtain the corresponding reconstructed data;

[0012] The reconstructed data is divided into multiple data subsets; a classification algorithm stored in the algorithm warehouse is applied to each subset; the optimal algorithm is selected for each data subset by measuring the evaluation index, and the algorithm is labeled; a database is built based on the labeled data.

[0013] The network is reconstructed based on the trained data. The data stored in the database and the data of the object to be predicted are encoded to obtain the corresponding data codes. Based on the corresponding data codes, the two are compared by the nearest neighbor algorithm, and the classification algorithm with the highest data fit is selected. The classification algorithm is then applied to obtain the behavioral tendency prediction result of the object to be evaluated.

[0014] A second aspect of the present invention provides a behavioral risk tendency prediction system based on the fusion of subjective and objective information, comprising:

[0015] The data processing module is used to acquire subjective and objective information of the object to be evaluated and process it to obtain the integrated subjective and objective information of the object to be evaluated.

[0016] The encoding module is used to combine subjective and objective information with trained data to reconstruct the network and obtain the corresponding reconstructed data.

[0017] The database construction module is used to divide the reconstructed data into multiple data subsets; apply the classification algorithm stored in the algorithm warehouse to each data subset, select the optimal algorithm for each data by measuring the evaluation index, and label the algorithm; and build the database based on the labeled data.

[0018] The rating assessment module is used to reconstruct the network based on the trained data. It encodes the data stored in the database and the data of the object to be predicted to obtain the corresponding data codes. Based on the corresponding data codes, the nearest neighbor algorithm is used to compare the two and select the classification algorithm with the highest data fit. The classification algorithm is then applied to obtain the behavioral tendency prediction result of the object to be evaluated.

[0019] A third aspect of the present invention provides a computer-readable storage medium.

[0020] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in a behavioral risk tendency prediction method based on a fusion representation of subjective and objective information as described in the first aspect.

[0021] A fourth aspect of the present invention provides a computer device.

[0022] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the steps of a behavioral risk tendency prediction method based on the fusion of subjective and objective information as described in the first aspect.

[0023] Compared with the prior art, the beneficial effects of the present invention are:

[0024] 1. This invention adopts an open model, which allows for selection of the preprocessing of raw data, data reconstruction, algorithm storage, data comparison, and algorithm selection methods. It can be applied to different evaluation objects, cope with different environments, and different subjective and objective information, thereby improving the generalization ability of the algorithm.

[0025] 2. This invention uses a combination of subjective and objective information representation for behavior prediction, which reduces the impact of human factors on the evaluation results and improves the accuracy of behavior prediction.

[0026] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0027] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0028] Figure 1 This is a schematic diagram of the overall system according to Embodiment 1 of the present invention;

[0029] Figure 2 This is a structural diagram of the overall frame of the self-encoder according to Embodiment 1 of the present invention;

[0030] Figure 3 This is a flowchart of the point-to-point ranking clustering algorithm of Embodiment 1 of the present invention;

[0031] Figure 4 This is a flowchart of the random forest algorithm according to Embodiment 1 of the present invention. Detailed Implementation

[0032] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0033] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0034] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0035] In view of the shortcomings of the prior art mentioned in the background, the present invention adopts an open model. Based on the traditional behavior prediction method, the present invention uses the method of computer deep learning to process the acquired object structured basic information and periodic assessment information to obtain a feature vector representing the fusion of subjective and objective information.

[0036] The subjective and objective information is input into the autoencoder to obtain data encoding and a trained encoder. A classification algorithm from the algorithm repository is applied to the data represented by the fused subjective and objective information. An appropriate algorithm is selected for each data point based on evaluation metrics, and the algorithm is labeled. A database is built based on this labeled data. For behavior prediction, the trained encoder is first used to encode the data stored in the database and the test data to obtain database data encoding and test data encoding. The nearest neighbor algorithm is used to compare the two and select the classification algorithm most suitable for the test data. The selected classification algorithm is then applied to obtain the behavior prediction result.

[0037] Example 1

[0038] like Figure 1 As shown, this embodiment provides a method for predicting behavioral risk tendencies based on the fusion of subjective and objective information, including the following steps:

[0039] Step 1: Obtain subjective and objective information about the object to be evaluated;

[0040] Among them, subjective information mainly consists of periodic assessment data, such as quantitative evaluations of the assessed person's performance this month, questionnaire results, and other information.

[0041] The objective information mainly includes three aspects: first, social background data, such as the age, gender, education level, occupation, and marital status of the person being assessed; second, physiological data such as heart rate and body temperature obtained through wearable devices, and the location information of the person being assessed; and finally, video and image information obtained through surveillance. Social background data contains basic information about the person being assessed in the supervised location and their original social environment, providing clues for predicting the person's dangerous tendencies; physiological data is used to determine whether the person is in a state of anger, tension, etc.; and location and image information are used to determine the person's behavioral patterns and activity range, thereby revealing potential dangerous tendencies.

[0042] Step 2: Preprocess the subjective and objective information of the object to be evaluated, integrate the two types of information, and obtain a fused representation of the subjective and objective information of the object to be evaluated.

[0043] Step 3: Reconstruct the network based on the fusion representation of the subjective and objective information of the object to be evaluated and the trained data to obtain the corresponding reconstructed data.

[0044] The data reconstruction network used is an autoencoder, and its structure is as follows: Figure 2 As shown. The autoencoder consists of an encoder and a decoder. The training process of the autoencoder is as follows: the subjective and objective fusion information of the object to be evaluated is used as the input of the encoder, and the output is the data encoding; the data encoding is used as the input of the decoder, and the output is the reconstructed data. By minimizing the feature distance between the reconstructed data and the input data, the encoder is guided to output the data encoding.

[0045] The autoencoder consists of a structurally symmetrical encoder and decoder. For an input vector space X∈Φ composed of subjective and objective fusion information, the encoder obtains a data encoding vector space C∈Θ, and the decoder restores the data encoding to the input vector as much as possible. The autoencoder solves the mapping relationship f and r between the two to minimize the reconstruction error.

[0046] The mathematical expression is as follows:

[0047] f:Φ→Θ

[0048] r:Θ→Φ

[0049]

[0050] The encoder consists of a three-layer fully connected network: the first layer is the input layer, and the second and third layers are hidden layers. The number of nodes in the first layer is the same as the dimension of the input data. In this embodiment, the number of nodes in the second layer is set to 64, and the number of nodes in the third layer is set to 32.

[0051] The decoder consists of a three-layer fully connected network: the first and second layers are hidden layers, and the third layer is the output layer. In this embodiment, the number of nodes in the first layer is set to 32, the number of nodes in the second layer is 64, and the number of nodes in the third layer is the same as that in the input layer.

[0052] This study employs an autoencoder algorithm from deep learning, using the subjective and objective information of the evaluation subject as input to the encoder. Through multiple iterative training sessions on a computer, it extracts general feature information of the evaluated individual, and then uses these extracted features for data comparison. When the database is large, this method can significantly improve the efficiency of behavior prediction and has high practicality. Integrating the algorithm into specific behavior prediction instruments allows for behavior prediction operations accessible to ordinary people, avoiding the waste of significant human resources and problems such as flawed evaluation logic due to insufficient prior knowledge.

[0053] Step 4: Apply a clustering algorithm to the dataset composed of data encoding to divide the dataset into multiple data subsets.

[0054] In this embodiment, the clustering algorithm used is the K-means algorithm, and its specific process is as follows:

[0055] (a) Set the number of data groups to K=5, and randomly select 5 initial cluster centers from all the subjective and objective information fusion data.

[0056] (b) Calculate the cosine distance between other sample data and the cluster center, and assign all sample data to the corresponding cluster according to the principle of minimum distance.

[0057] (c) Update the cluster centers using the mean of the sample data for each cluster. For a cluster of sample data s1, s2, ..., s n ∈S, representing the updated cluster center c(a1,a2,…,a…). k )for:

[0058]

[0059] Among them, a k This represents the k-th feature value of the cluster center. This represents the k-th feature value of the i-th sample.

[0060] (d) Repeat steps (b) and (c) until the change in each dimension of the cluster center is less than 0.1.

[0061] Step 5: Apply the classification algorithm stored in the algorithm warehouse to each subset of data, select the optimal algorithm for each data point by measuring the evaluation metrics, and label the algorithm. Build a database based on the labeled data.

[0062] The classification algorithm repository stores various classification algorithms, including the point-sorting clustering algorithm, the Naive Bayes algorithm, the C4.5 algorithm, and the random forest algorithm.

[0063] The aforementioned point-sorting clustering algorithm is a density-based clustering algorithm. This algorithm does not explicitly provide classification results, but rather obtains an ordered list by sorting the objects in the dataset, thereby generating a decision graph. Finally, it sets a minimum threshold that meets the conditions based on the number of levels of behavioral risk tendency, thus achieving the prediction of behavioral risk tendency.

[0064] The main concepts involved in the point-ranking clustering algorithm are as follows:

[0065] (a) ε-neighborhood: For a given sample p∈X, the ε-neighborhood represents the set of samples in the sample set X whose distance from p is not greater than ε, and the number of elements in this set is N. ε (p) can be represented as:

[0066] N ε (p)={x i ∈X|distance(x i ,p)≤ε}

[0067] (b) Core Object: A preset sample threshold M is defined. For any sample p∈X in the sample set, if N satisfies ε (p)≥M, then x j It is the core object.

[0068] (c) Core distance: The minimum distance that makes a sample p∈X exactly a core object.

[0069] (d) Reachability distance: for samples p, x i For any x, the reachable distance between the two is defined as the larger of the Euclidean distance and the core distance.

[0070] (e) Density direct access: For sample p, x i ∈X, if x i If x is within the ε-neighborhood of p and p is a core object, then x is called x. i It reaches the density directly at p.

[0071] The flowchart of the point-to-point ranking clustering algorithm is as follows: Figure 3 As shown, the specific process is as follows:

[0072] (a) Based on the dataset, set ε and M, initialize the dataset X consisting of subjective and objective information fusion representation data, and mark the data samples in it as unvisited.

[0073] (b) Randomly select a data sample p from dataset X.

[0074] (c) Determine whether data sample p is a core object, that is, determine whether the number of data samples contained in the ε-neighborhood of p satisfies N. ε (p)≥M

[0075] (d) If p is a core object, mark p as visited and add it to the processed sequence O. Extract all samples that are directly accessible from sample p to obtain the sample set N. Sort all data in N according to reachability distance and add them to the unprocessed sequence Q.

[0076] (e) Take the first sample in Q that has the smallest distance to the sample q, mark q as visited and put it into the processed sequence O.

[0077] (f) Let p = q, and repeat the operations c, d, e until Q is an empty set.

[0078] (g) Repeat operations b, c, d, e, f until X is an empty set, at which point the algorithm ends.

[0079] The Naive Bayes algorithm uses the features of the evaluation object across various dimensions of subjective and objective information as prior conditions to calculate the probability that the evaluation object belongs to various behavioral risk tendencies, and selects the term with the highest probability as the final evaluation result. The specific process is as follows:

[0080] (a) For a piece of information x that combines subjective and objective elements, we have the expression x = {a1, a2, a3, ..., a...} n}, the set of behavioral risk tendency categories C = {y1, y2, y3, ..., y m ,}. Where a i To evaluate the characteristic attributes of the object by integrating subjective and objective information, y j This indicates a tendency toward dangerous behavior. 'n' represents the feature dimension of each piece of information combining subjective and objective factors, and 'm' represents the number of categories of dangerous behavior tendencies.

[0081] (b) Calculate P(y1|x), P(y2|x), P(y3|x), ..., P(y m |x) represents the probability that the assessed object belongs to each behavioral risk tendency. Assuming that the characteristics are independent, we can obtain the following from Bayes' theorem:

[0082]

[0083] (c) Determine P(y) that satisfies condition k |x)=max{P(y1|x),P(y2|x),…,P(y m The k value of |x)} determines the risk level of the assessed object's behavior as k.

[0084] The C4.5 algorithm is a type of decision tree algorithm. Based on the information gain ratio, it constructs a decision tree that can select relatively important features from all dimensions of subjective and objective information. When using the C4.5 algorithm for behavior prediction, the main task is the construction of the decision tree.

[0085] A decision tree is a tree-like structure that performs feature-based classification. It groups labeled samples multiple times by calculating predefined metrics. Each node represents a feature attribute used for classification, the corresponding branch represents the grouping result using this feature attribute, and the leaf nodes store a behavioral risk tendency category. C4.5 uses information gain ratio as the metric for constructing a decision tree. For a set of subjective and objective information fusion sample data D, if the behavioral risk tendency categories it contains can be divided into x1, x2, ..., x... k These k classes, each with a probability of appearing in D, are p1, p2, ..., p. k Then the entropy of this set of data D is:

[0086]

[0087] If we group D based on the feature attribute 'a', and 'a' has v possible values, then D can be divided into D1, D2, ..., D... v We obtain v subsets of data. The information gain of this grouping method is:

[0088]

[0089] The corresponding information gain rate is:

[0090]

[0091] If the feature dimension of this set of subjective and objective integrated information sample data D is N, and the corresponding feature attributes are (a1, a2, ..., a...), then... N ), calculate the information gain ratio {GainRatio(D,a1), GainRatio(D,a2), ..., GainRatio(D,a3)} for each feature attribute. N If the maximum value is GainRatio(D,a) l If ), then choose a. l This is used to partition the sample data D and obtain the corresponding subsets. Iterating through the above steps generates a decision tree. For unlabeled subjective-objective fusion test data, starting from the root node, multiple feature attribute judgments are performed along the decision tree path to find the corresponding leaf node. The behavioral risk tendency category stored in the leaf node is the evaluation result.

[0092] The random forest algorithm is an ensemble learning method with decision trees as classifiers. Based on the output results of all decision trees, a majority vote is conducted on the behavioral risk tendency of the evaluation object, thereby obtaining the final evaluation result. Its basic process is as Figure 4 shown:

[0093] For the set D of subjective and objective fusion information data, random resampling is performed to obtain data encoding subsets S(1), S(2), …, S(k), and k decision trees are generated from the k sub-datasets. When making behavioral predictions, the generated model is applied to the test samples to obtain k behavioral prediction results, and finally the behavioral risk tendency of the evaluation object is determined by majority voting.

[0094] Random resampling is used for the data encoding set to avoid generating the same decision tree. Suppose there are M pieces of data in the data encoding set output by the encoder, and the feature dimension of each piece of data is N. When performing resampling, each time m (m < M) pieces of data are randomly drawn with replacement from the dataset of size M, and then n (n < N) dimensions are randomly selected from the N-dimensional features of the m pieces of data to form the data encoding subset S. Repeating the above operation k times yields the dataset used to generate the decision tree.

[0095] To measure the classification effect of each algorithm, the evaluation metrics used are accuracy and micro F1. Apply all classification algorithms in the algorithm repository to each group of sample data, and calculate the accuracy and micro F1 of each algorithm.

[0096] The expression of the evaluation metric accuracy is as follows:

[0097]

[0098] Among them, T represents the number of correctly classified samples, and N represents the total sample size.

[0099] The expression of the evaluation metric micro F1 is as follows:<当进行行为预测时,对测试样本应用生成的模型,得到k个行为预测结果,最后由多数投票确定评估对象的行为危险倾向。

[0100]

[0101] Since the traditional F1 value is not applicable to multi-classification tasks, micro_F is used as the evaluation criterion. Among them, Precision micro and Recall micro respectively represent the total Precision and Recall of all classes, and the expressions are as follows:

[0102]

[0103]

[0104] Among them, n is the total number of classes, and i represents the i-th class of behavioral risk tendency.

[0105] The classification algorithms in the algorithm repository are coded as follows: point sorting clustering algorithm is coded as 0, Naive Bayes algorithm as 1, decision tree algorithm as 2, and random forest algorithm as 3. Based on the accuracy and micro F1 score obtained by each classification algorithm after applying it to each group of data, the applicable algorithms for all subjective and objective fusion data are determined, the data are labeled with algorithms, and these data are integrated to obtain the constructed database.

[0106] Step 5: Behavior prediction specifically includes:

[0107] Step 501: Reconstruct the network based on the trained data and encode the subjective and objective fusion information data of the evaluation object and the data in the database.

[0108] Step 502: Perform data comparison. Use the nearest neighbor algorithm to calculate the Euclidean distance between the test data code and the data codes in the database. Select the N data that are closest to the test data code and conduct a majority vote. Select a classification algorithm from the algorithm warehouse based on the voting results.

[0109] For the test data encoding X(x1,x2,…,x…), n ) and database data encoding S(s1,s2,…,s n The Euclidean distance between the two is calculated using the following formula:

[0110]

[0111] Step 503: Apply the selected classification algorithm to the test data and output the behavior prediction results.

[0112] Example 2

[0113] This embodiment provides a behavioral risk tendency prediction system based on the fusion of subjective and objective information, characterized in that it includes:

[0114] The data processing module is used to acquire subjective and objective information of the object to be evaluated and process it to obtain the integrated subjective and objective information of the object to be evaluated.

[0115] The encoding module is used to combine subjective and objective information with trained data to reconstruct the network and obtain the corresponding reconstructed data.

[0116] The database construction module is used to divide the reconstructed data into multiple data subsets; apply the classification algorithm stored in the algorithm warehouse to each data subset, select the optimal algorithm for each data by measuring the evaluation index, and label the algorithm; and build the database based on the labeled data.

[0117] The rating assessment module is used to reconstruct the network based on the trained data. It encodes the data stored in the database and the data of the object to be predicted to obtain the corresponding data codes. Based on the corresponding data codes, the nearest neighbor algorithm is used to compare the two and select the classification algorithm with the highest data fit. The classification algorithm is then applied to obtain the behavioral tendency prediction result of the object to be evaluated.

[0118] Example 3

[0119] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in a behavioral risk tendency prediction method based on subjective and objective information fusion representation as described in Embodiment 1.

[0120] Example 4

[0121] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the behavioral risk tendency prediction method based on subjective and objective information fusion representation as described in Embodiment 1.

[0122] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0123] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0124] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0126] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0127] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting behavioral risk tendencies based on the fusion of subjective and objective information, characterized in that, Includes the following steps: The process involves acquiring subjective and objective information about the object to be predicted, and then processing this information to obtain a fusion of subjective and objective information. Subjective information includes periodic assessment data. Objective information includes three aspects: social background data, physiological data obtained through wearable devices, and location information of the object being assessed, as well as video information obtained through monitoring. By combining subjective and objective fusion information with a trained data reconstruction network, corresponding reconstructed data is obtained. The data reconstruction network is an autoencoder, which consists of an encoder and a decoder. The training process of the autoencoder is as follows: the subjective and objective fusion information of the object to be evaluated is used as the input of the encoder, and the output is the data encoding; the data encoding is used as the input of the decoder, and the output is the reconstructed data. By minimizing the feature distance between the reconstructed data and the input data, the encoder is guided to output the data encoding. The autoencoder consists of a structurally symmetrical encoder and decoder. For an input vector space X∈Φ composed of subjective and objective fusion information, the encoder obtains a data encoding vector space C∈Θ, and the decoder restores the data encoding to the input vector as much as possible. The autoencoder solves the mapping relationship f and r between the two to minimize the reconstruction error, which can be mathematically expressed as follows: ， Where f represents the encoder mapping function and r represents the decoder mapping function. Represents the original input feature space. This represents the data encoding vector space output by the encoder. This represents the input vector containing the combined subjective and objective information of a single sample. This represents the hidden layer coding vector obtained by compressing sample X using the encoder. The reconstructed data is divided into multiple data subsets; classification algorithms stored in the algorithm warehouse are applied to each data subset, and the optimal algorithm is selected for each data point based on evaluation metrics. Algorithm labels are then assigned, and a database is constructed based on the labeled data. This includes: the classification algorithms in the algorithm warehouse include point-ranking clustering, Naive Bayes, C4.5, and Random Forest algorithms; based on the accuracy and micro F1 score obtained after applying each classification algorithm to each group of data, the applicable algorithms for all subjective and objective fusion data are determined, the data are labeled with algorithms, and these data are integrated to obtain the constructed database. Based on the trained data, the network is reconstructed, and the data stored in the database and the data of the object to be predicted are encoded to obtain corresponding data codes. Based on the corresponding data codes, the nearest neighbor algorithm is used to compare the two, and the classification algorithm with the highest data fit is selected. The classification algorithm is then applied to obtain the behavioral tendency prediction result of the object to be evaluated. The process includes: performing data comparison, using the nearest neighbor algorithm to calculate the Euclidean distance between the test data code and the data code in the database, selecting the top N data with the closest distance to the test data code and conducting a majority vote, selecting a classification algorithm from the algorithm warehouse based on the voting result, applying the selected classification algorithm to the test data, and outputting the behavioral prediction result.

2. The behavioral risk tendency prediction method based on subjective and objective information fusion as described in claim 1, characterized in that, When dividing the reconstructed data into multiple data subsets, the K-means algorithm is used.

3. The behavioral risk tendency prediction method based on subjective and objective information fusion as described in claim 1, characterized in that, The process of dividing the reconstructed data into multiple data subsets specifically includes: randomly selecting initial cluster centers, calculating the cosine distance between each data set and the cluster center, grouping the data according to the nearest distance principle, and calculating new cluster centers for each group; iteratively updating the cluster centers until they remain unchanged or fluctuate within a certain range.

4. A behavioral risk tendency prediction system based on the fusion of subjective and objective information, characterized in that, The method for predicting behavioral risk tendencies based on the fusion of subjective and objective information as described in any one of claims 1-3 includes: The data processing module is used to acquire subjective and objective information of the object to be evaluated and process it to obtain the integrated subjective and objective information of the object to be evaluated. The encoding module is used to combine subjective and objective information with trained data to reconstruct the network and obtain the corresponding reconstructed data. The database construction module is used to divide the reconstructed data into multiple data subsets; apply the classification algorithm stored in the algorithm warehouse to each data subset, select the optimal algorithm for each data by measuring the evaluation index, and label the algorithm; and build the database based on the labeled data. The rating assessment module is used to reconstruct the network based on the trained data. It encodes the data stored in the database and the data of the object to be predicted to obtain the corresponding data codes. Based on the corresponding data codes, the nearest neighbor algorithm is used to compare the two and select the classification algorithm with the highest data fit. The classification algorithm is then applied to obtain the behavioral tendency prediction result of the object to be evaluated.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the behavioral risk tendency prediction method based on the fusion of subjective and objective information as described in any one of claims 1-3.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the behavioral risk tendency prediction method based on the fusion of subjective and objective information as described in any one of claims 1-3.

Images