Intervention result prediction model generation method, prediction method and related equipment
A decision tree-based method uses feedback information under multiple intervention conditions to predict individual intervention outcomes, addressing the challenge of determining individual-specific strategies by constructing a model that accurately forecasts intervention results.
Patent Information
- Application Number
- CN202310747677.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-21
- Publication Date
- 2024-12-31
AI Technical Summary
Existing methods fail to determine individual-specific intervention strategies due to the inability to observe changes under different interventions in group experiments, making it impossible to predict intervention outcomes for individuals.
A method is developed to construct a decision tree using information gain based on differences in feedback information under multiple intervention conditions, dividing nodes using sample features to generate child nodes and determining prediction results for each leaf node, allowing prediction of intervention outcomes for individuals.
The method enables efficient determination of intervention strategies for individuals by predicting intervention outcomes based on overall intervention results, improving the accuracy and efficiency of intervention prediction.
Smart Images
Figure CN119226941A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of data processing, and in particular to a method for generating an intervention outcome prediction model, a prediction method, and related equipment. Background Art
[0002] Data-driven artificial intelligence and its related technologies have played a huge role in all walks of life and brought high value. In addition to the commonly used technologies, how to measure the impact of intervention conditions on specific individuals has become one of the important technical difficulties.
[0003] In related technologies, different interventions can be randomly given to each group sample through random experiments, so as to obtain the degree of influence of the intervention conditions on the group. Summary of the invention
[0004] After analysis, the inventors found that since each individual in the experimental group could only receive one intervention, the changes under different interventions could not be observed. Therefore, it was impossible to discover the intervention effect on the individual from the results of a single group experiment, and it was impossible to determine the intervention strategy for the individual.
[0005] A technical problem to be solved by the embodiments of the present disclosure is: how to efficiently determine an intervention strategy for an individual.
[0006] According to a first aspect of some embodiments of the present disclosure, a method for generating an intervention outcome prediction model is provided, comprising: using samples, and differences in feedback information of samples under multiple intervention conditions, as information gain of sample features, to construct a decision tree using information gain, wherein non-leaf nodes in the decision tree are divided using sample features to generate child nodes; for each leaf node of the decision tree, determining a prediction result corresponding to the leaf node based on feedback values of the samples corresponding to the leaf node under multiple intervention conditions, so that the decision tree can be used to predict intervention outcomes for multiple intervention conditions.
[0007] In some embodiments, information gain is determined in the process of building a decision tree based on the difference in feedback information of different sample sets under multiple intervention conditions, and the different sample sets are generated by pre-dividing the samples using their characteristics.
[0008] In some embodiments, the feedback information of each sample set is: when multiple intervention conditions are applied to the samples in the sample set respectively and the mean of the sample feedback value corresponding to each intervention condition is determined, the sum of the differences between the means of the feedback values corresponding to the two intervention conditions in each intervention combination, wherein the intervention combination is obtained by traversing the combination of two different intervention conditions among the multiple intervention conditions.
[0009] In some embodiments, the samples are divided into a plurality of groups corresponding to a plurality of intervention conditions, and the corresponding intervention condition is applied to the samples in each group.
[0010] In some embodiments, samples and differences in feedback information of samples under multiple intervention conditions are used as information gains of features of samples, and a decision tree is constructed using information gain, including: using differences in feedback information of samples corresponding to the node to be processed under multiple intervention conditions to calculate the information gains of each current candidate feature, wherein the node to be processed is the current leaf node of the decision tree, and the current candidate feature is a feature that has not yet been used to divide the nodes in the decision tree; selecting candidate features for division based on the information gain of each candidate feature; and using the candidate features for division to divide the node to be processed and its corresponding samples to generate child nodes of the node to be processed and determine the samples corresponding to each child node.
[0011] In some embodiments, the information gain of each current candidate feature is calculated by using the difference in feedback information of the samples corresponding to the node to be processed under multiple intervention conditions, including: for each candidate feature: using the alternative feature to pre-divide the samples corresponding to the node to be processed, and generate a first sample set and a second sample set corresponding to the alternative feature; determining the difference in feedback information between the first sample set and the second sample set under multiple intervention conditions as the information gain of the alternative feature.
[0012] In some embodiments, determining the difference in feedback information between a first sample set and a second sample set under multiple intervention conditions as the information gain of an alternative feature includes: for each set in the first sample set and the second sample set corresponding to the alternative feature, traversing the combination of the two intervention conditions in multiple intervention conditions, calculating the difference in the mean of the feedback values of the samples in the set under the two intervention conditions in each combination, and calculating the sum of these differences as the feedback information of the set; determining the information gain of the alternative feature based on the difference in feedback information between the first sample set and the second sample set.
[0013] In some embodiments, the number of samples corresponding to the node to be processed is greater than the sample threshold.
[0014] In some embodiments, when the depth of the decision tree reaches a depth threshold, the construction of the decision tree is completed.
[0015] In some embodiments, determining the prediction result corresponding to the leaf node based on the feedback values of the samples corresponding to the leaf node under multiple intervention conditions includes: for each intervention condition, determining the mean of the feedback values of the samples corresponding to the leaf node under the intervention conditions as the expected value under the intervention conditions corresponding to the leaf node; and determining the prediction result corresponding to the leaf node based on the expected value corresponding to the leaf node under each intervention condition.
[0016] In some embodiments, determining the prediction result corresponding to the leaf node according to the expected value corresponding to the leaf node under each intervention condition includes: determining the expected value corresponding to the leaf node under each intervention condition as the prediction result corresponding to the leaf node.
[0017] In some embodiments, determining the prediction result corresponding to the leaf node according to the expected value corresponding to the leaf node under each intervention condition includes:
[0018] For each intervention condition, the expected value under the intervention condition corresponding to the leaf node is compared with the corresponding intervention result threshold to determine the effectiveness information of the intervention condition;
[0019] The effectiveness information of each intervention condition is determined as the prediction result corresponding to the leaf node.
[0020] In some embodiments, the plurality of intervention conditions comprises more than 3 intervention conditions.
[0021] In some embodiments, the sample is a biological sample, the intervention condition includes at least one of a drug type or a drug dosage, and the feedback information represents the reaction of the biological sample after the intervention condition is applied; or, the sample is a user of the application, the intervention condition includes at least one of a resource or information delivered to the user, and the feedback information represents the user's interaction with the application after the intervention condition is applied.
[0022] According to a second aspect of some embodiments of the present disclosure, a method for predicting intervention results is provided, comprising: inputting a sample to be tested into a decision tree, wherein the decision tree is generated by using any of the aforementioned methods for generating an intervention result prediction model; determining a leaf node to which the sample to be tested belongs in the decision tree; and determining the intervention results of the sample to be tested under multiple intervention conditions based on the prediction results corresponding to the leaf nodes.
[0023] According to a third aspect of some embodiments of the present disclosure, there is provided a device for generating an intervention outcome prediction model, comprising: a construction module, configured to utilize samples, and differences in feedback information of samples under multiple intervention conditions as information gain of sample features, to construct a decision tree, wherein non-leaf nodes in the decision tree are divided using sample features to generate child nodes; and a determination module, configured to determine, for each leaf node of the decision tree, a prediction result corresponding to a leaf node according to feedback values of the samples corresponding to the leaf node under multiple intervention conditions.
[0024] According to a fourth aspect of some embodiments of the present disclosure, there is provided a device for generating an intervention outcome prediction model, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute any of the aforementioned methods for generating an intervention outcome prediction model based on instructions stored in the memory.
[0025] According to a fifth aspect of some embodiments of the present disclosure, a system for predicting intervention results is provided, comprising: a device for generating any one of the aforementioned intervention result prediction models; and a prediction device, configured to input a sample to be tested into a decision tree, wherein the decision tree is generated by the device for generating the intervention result prediction model; determine the leaf node to which the sample to be tested belongs in the decision tree; and determine the intervention results of the sample to be tested under multiple intervention conditions based on the prediction results corresponding to the leaf nodes.
[0026] According to a sixth aspect of some embodiments of the present disclosure, there is provided a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the aforementioned methods for generating an intervention outcome prediction model.
[0027] Some embodiments of the above invention have the following advantages or beneficial effects. The present disclosure can construct a decision tree using the difference in feedback information of samples under multiple intervention conditions, so that the decision tree can predict the feedback results of the samples under multiple intervention conditions. Therefore, the embodiments of the present disclosure can use the overall intervention results of the samples to determine a model for predicting the intervention results of individuals, so that the intervention strategy for individuals can be determined efficiently.
[0028] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0030] Figure 1 A flowchart of a method for generating an intervention outcome prediction model according to some embodiments of the present disclosure is shown.
[0031] Figure 2 A schematic flow chart of a method for constructing a decision tree according to some embodiments of the present disclosure is shown.
[0032] Figure 3 A flowchart of a method for determining a prediction result of a leaf node according to some embodiments of the present disclosure is shown.
[0033] Figure 4 A schematic flowchart of a prediction method according to some embodiments of the present disclosure is shown.
[0034] Figure 5 A schematic diagram of the structure of a device for generating an intervention outcome prediction model according to some embodiments of the present disclosure is shown.
[0035] Figure 6 A schematic diagram of the structure of a system for predicting intervention results according to some embodiments of the present disclosure is shown.
[0036] Figure 7 A schematic structural diagram of a device for generating an intervention outcome prediction model according to other embodiments of the present disclosure is shown.
[0037] Figure 8 A schematic structural diagram of a device for generating an intervention outcome prediction model according to some further embodiments of the present disclosure is shown. DETAILED DESCRIPTION
[0038] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present disclosure and its application or use. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.
[0039] Unless specifically stated otherwise, the relative arrangement of components and steps, the numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present disclosure.
[0040] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.
[0041] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.
[0042] In all examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.
[0043] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.
[0044] Figure 1 FIG. 1 is a flow chart showing a method for generating an intervention outcome prediction model according to some embodiments of the present disclosure. Figure 1As shown, the method for generating the intervention result prediction model in this embodiment includes steps S102 to S104.
[0045] In step S102, the samples and the differences in feedback information of the samples under multiple intervention conditions are used as information gain of the characteristics of the samples, so as to construct a decision tree using information gain, wherein non-leaf nodes in the decision tree are divided using the characteristics of the samples to generate child nodes.
[0046] The sample is the object of intervention in the intervention experiment, such as biological samples, users, etc. Each sample has multiple features, which are used to indicate the attributes, labels, categories, behaviors and other information of the sample. The samples used in building a decision tree and using the decision tree for prediction can have the same features, but different samples can have different feature values. The identification, features and other information of the sample can be stored in the database as a record and can be read from the database when needed.
[0047] In some embodiments, the sample is a biological sample, the intervention condition includes at least one of a drug type or a drug dosage, and the feedback information indicates the reaction of the biological sample after the intervention condition is applied; or the sample is a user of the application, the intervention condition includes at least one of a resource or information delivered to the user, and the feedback information indicates the user's interaction with the application after the intervention condition is applied. As required, the embodiments of the present disclosure can also be applied to other technical fields, which will not be described in detail here.
[0048] In some embodiments, the samples are divided into a plurality of groups corresponding to a plurality of intervention conditions, and the samples in each group are subjected to corresponding intervention conditions. Thus, for each sample, only one experiment needs to be performed, that is, one intervention condition is applied. After the intervention, the feedback value of the sample can be recorded. The direct feedback result of each sample can be a numerical value or other types of information. If it is other types of information, it can be quantified and expressed as a numerical value, which facilitates the subsequent calculation process. The feedback value of the sample can be recorded in the data record corresponding to the sample in the database.
[0049] By using the characteristics of the samples, we can determine the partitioning conditions of the samples corresponding to the non-leaf nodes in the decision tree, and determine the child nodes corresponding to the samples according to the conditions met by the samples. For example, if the partitioning condition corresponding to a node is whether the age is greater than 30 years old, then for a user sample corresponding to the node, we can determine whether it is the left child node or the right child node of the node according to the partitioning condition.
[0050] Information gain is an important basis for decision tree construction. Information gain indicates how much information a feature can bring to the classification system. The more information a feature brings, the more important the feature is, and accordingly, the greater its information gain is.
[0051] The embodiment of the present disclosure utilizes a new method of calculating information gain, that is, taking the difference in feedback information of a sample under multiple intervention conditions as the information gain of the sample's features. Therefore, the decision tree can distinguish samples with large differences in feedback to multiple intervention conditions as much as possible, thereby effectively predicting the intervention results.
[0052] In some embodiments, information gain is determined in the process of constructing a decision tree based on the difference in feedback information of different sample sets under multiple intervention conditions, and the different sample sets are generated by pre-dividing using the features of the samples. For example, when determining which feature is currently used to divide a node to generate a child node, each candidate feature is used in turn for pre-division, and the candidate feature is a feature that is not currently used for the actual division of the decision tree. For each candidate feature, the information gain corresponding to the candidate feature can be determined based on the difference in feedback information of different sets generated by the pre-division under multiple intervention conditions.
[0053] In some embodiments, the feedback information of each set is determined in the following manner. First, multiple intervention conditions are applied to the samples in the sample set, and the mean of the sample feedback value corresponding to each intervention condition is determined; then, the combination of two different intervention conditions in the multiple intervention conditions is obtained by traversing, that is, all combinations of intervention conditions of 2 in the multiple intervention conditions are generated, and then the difference between the mean values of the feedback values corresponding to the two intervention conditions in each combination is calculated; finally, the sum of these differences is calculated as the feedback information of the set. Thus, the feedback information of the set can reflect the difference in sample feedback under any two intervention conditions.
[0054] In this way, when the number of intervention conditions is more than 3, the difference between every two intervention conditions can be calculated respectively, so that the feedback information can reflect the difference information between the various intervention conditions.
[0055] In step S104, for each leaf node of the decision tree, the prediction result corresponding to the leaf node is determined according to the feedback value of the sample corresponding to the leaf node under multiple intervention conditions, so that the decision tree is used to predict the intervention results of multiple intervention conditions.
[0056] In the related art, each leaf node of a decision tree usually corresponds to a category. In the decision tree constructed by the embodiments of the present disclosure, the leaf node can also be regarded as corresponding to a category, but is further endowed with more meanings. That is, according to the decision tree construction stage, the performance of the samples belonging to the leaf node on each intervention condition is used as the "category" of the leaf node. Thus, in the use stage of the decision tree, the leaf node to which the tested sample belongs can be used to efficiently predict its feedback results under the application of each intervention condition.
[0057] Through the above embodiments, the present disclosure can construct a decision tree using the differences in feedback information of samples under multiple intervention conditions, so that the decision tree can predict the feedback results of samples under multiple intervention conditions. Therefore, the embodiments of the present disclosure can use the overall intervention results of the samples to determine a model for predicting the intervention results of individuals, thereby being able to efficiently determine the intervention strategy for individuals.
[0058] Reference below Figure 2 , a method of constructing a decision tree is described exemplarily.
[0059] Figure 2 FIG. 1 is a flow chart showing a method for constructing a decision tree according to some embodiments of the present disclosure. Figure 2 As shown, the decision tree construction method of this embodiment includes S202 to S208.
[0060] Before executing each step, the sample set is divided into multiple groups according to the number of intervention conditions. The corresponding intervention conditions are applied to the samples in each group, and the feedback value of each sample is recorded.
[0061] In step S202, a root node of a decision tree is generated, wherein the root node corresponds to all samples used for training.
[0062] Then, steps S204 to S208 are repeated until the stop condition is reached, wherein the node to be processed is the current leaf node of the decision tree, and the current candidate feature is a feature that has not been used to divide the nodes in the decision tree. In some embodiments, the number of samples corresponding to the node to be processed is greater than the sample threshold, thereby avoiding overfitting of the model and improving the accuracy of decision tree prediction.
[0063] In some embodiments, the stopping condition is that the depth of the decision tree reaches a depth threshold; or, there is currently no node whose corresponding sample quantity is greater than the sample threshold.
[0064] In step S204, the information gain of each current candidate feature is calculated using the difference in feedback information of the samples corresponding to the node to be processed under multiple intervention conditions.
[0065] In some embodiments, for each candidate feature, the following processing is performed. First, the samples corresponding to the node to be processed are pre-divided using the candidate feature to generate a first sample set and a second sample set corresponding to the candidate feature; then, the difference in feedback information between the first sample set and the second sample set under multiple intervention conditions is determined as the information gain of the candidate feature. Thus, the result of dividing the node to be processed using various features can be simulated, so as to determine the appropriate features for division, thereby improving the accuracy of decision tree construction.
[0066] In some embodiments, the information gain may be determined in the following manner: For each set in the first sample set and the second sample set corresponding to the candidate feature, traverse the combination of two intervention conditions in multiple intervention conditions, calculate the difference between the mean values of the feedback values of the samples in the set under the two intervention conditions in each combination, and calculate the sum of these differences as the feedback information of the set; determine the information gain of the candidate feature according to the difference between the feedback information of the first sample set and the second sample set.
[0067] For example, feature 1 "whether the age is greater than 30" is used to divide the samples corresponding to a node to be processed into set 1 and set 2. Assuming there are 4 intervention conditions, the samples in set 1 and set 2 can be divided into 4 groups according to the intervention conditions applied, and the mean of the feedback value of the samples in each group is calculated.
[0068] Then, for set 1, all combinations of intervention conditions including 2 intervention conditions are generated, such as {group 1, group 2}, {group 1, group 3}, {group 1, group 4}, {group 2, group 3}, {group 2, group 4}, {group 3, group 4}, and the difference of the means corresponding to each combination is calculated, and then these differences are summed up as the feedback information of set 1. Set 2 is calculated in a similar way. Finally, the absolute value of the difference between the feedback information of set 1 and set 2 is taken as the information gain of feature 1.
[0069] In step S206, a candidate feature for segmentation is selected according to the information gain of each candidate feature. For example, the candidate feature with the largest information gain is selected.
[0070] In step S208, the node to be processed and its corresponding samples are divided using the candidate features for division to generate child nodes of the node to be processed and determine the samples corresponding to each child node.
[0071] The above embodiment uses the difference in feedback information of samples under multiple intervention conditions as the information gain of the sample features. Therefore, the decision tree can distinguish samples with large differences in feedback on multiple intervention conditions as much as possible, thereby effectively predicting the intervention results.
[0072] Reference below Figure 3 An embodiment of determining a prediction result corresponding to a leaf node is described.
[0073] Figure 3 FIG. 1 is a flow chart showing a method for determining a prediction result of a leaf node according to some embodiments of the present disclosure. Figure 3 As shown, the prediction result determination method of this embodiment includes steps S302 to S304.
[0074] In step S302, for each intervention condition, the mean of the feedback values of the samples corresponding to the leaf node under the intervention condition is determined as the expected value under the intervention condition corresponding to the leaf node.
[0075] In some embodiments, the expected value corresponding to the leaf node under each intervention condition is determined as the prediction result corresponding to the leaf node. This method can obtain a quantitative and more refined prediction result.
[0076] In some embodiments, for each intervention condition, the expected value under the intervention condition corresponding to the leaf node is compared with the corresponding intervention result threshold to determine the validity information of the intervention condition; the validity information of each intervention condition is determined as the prediction result corresponding to the leaf node. This method can obtain qualitative prediction results with higher reliability.
[0077] For example, for the samples corresponding to a leaf node, the mean feedback value for intervention condition 1 is 3.5, the mean feedback value for intervention condition 2 is 4.1, the mean feedback value for intervention condition 3 is 0.8, and the mean feedback value for intervention condition 4 is -2.
[0078] One way is to use {intervention condition 1: 3.5; intervention condition 2: 4.1; intervention condition 3: 0.8; intervention condition 4: -2} as the prediction result corresponding to the leaf node.
[0079] Another way is to set the threshold of intervention condition 1 to 3, and it is valid if it is greater than the threshold, the threshold of intervention condition 2 to 4, and it is valid if it is greater than the threshold, the threshold of intervention condition 3 to 1, and it is valid if it is less than the threshold, and the threshold of intervention condition 4 to 0, and it is valid if it is less than the threshold, then {Intervention condition 1: valid; Intervention condition 2: valid; Intervention condition 3: valid; Intervention condition 4: valid} can be used.
[0080] In step S304, the prediction result corresponding to the leaf node is determined according to the expected value corresponding to the leaf node under each intervention condition.
[0081] It can be seen from the above embodiments that by applying an intervention condition to each sample and constructing a decision tree model, the obtained model can predict the feedback results of the individual under each intervention condition. Therefore, the feedback results of the individual under each intervention condition can be obtained efficiently.
[0082] Reference below Figure 4 An embodiment of the prediction method of the present disclosure is described.
[0083] Figure 4 FIG. 4 is a flow chart showing a prediction method according to some embodiments of the present disclosure. Figure 4As shown, the prediction method of this embodiment includes steps S402 to S406.
[0084] In step S402, the sample to be tested is input into a decision tree, wherein the decision tree is constructed by using the method of any one of the aforementioned embodiments.
[0085] In step S404, the leaf node to which the sample to be tested belongs in the decision tree is determined.
[0086] For example, starting from the root node, according to the division condition corresponding to each node, determine which child node the sample to be tested should correspond to, until it corresponds to a leaf node.
[0087] In step S406, the intervention results of the sample to be tested under multiple intervention conditions are determined according to the prediction results corresponding to the leaf nodes.
[0088] For example, if the prediction result of the leaf node is {intervention condition 1: valid; intervention condition 2: valid; intervention condition 3: invalid; intervention condition 4: invalid}, it means that intervention conditions 1 and 2 are valid for the sample to be tested, and intervention conditions 3 and 4 are invalid for the sample to be tested.
[0089] Through the above embodiments, the experimental results can be obtained without conducting experiments on the samples to be tested, and the prediction of the intervention results of multiple intervention conditions can be obtained at one time. Therefore, the embodiments of the present disclosure can improve the efficiency of information processing.
[0090] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security and national security.
[0091] Reference below Figure 5 An embodiment of a device for generating an intervention outcome prediction model disclosed herein is described.
[0092] Figure 5 FIG. 2 shows a schematic diagram of a structure of a device for generating an intervention result prediction model according to some embodiments of the present disclosure. Figure 5As shown, the generation device 500 of the intervention result prediction model of this embodiment includes: a construction module 5100, which is configured to use the difference between the sample and the feedback information of the sample under multiple intervention conditions as the information gain of the sample's characteristics to construct a decision tree, wherein the non-leaf nodes in the decision tree are divided using the characteristics of the sample to generate child nodes; a determination module 5200, which is configured to determine the prediction result corresponding to each leaf node of the decision tree according to the feedback value of the sample corresponding to the leaf node under multiple intervention conditions.
[0093] In some embodiments, information gain is determined in the process of building a decision tree based on the difference in feedback information of different sample sets under multiple intervention conditions, and the different sample sets are generated by pre-dividing the samples using their characteristics.
[0094] In some embodiments, the feedback information of each sample set is: when multiple intervention conditions are applied to the samples in the sample set respectively and the mean of the sample feedback value corresponding to each intervention condition is determined, the sum of the differences between the means of the feedback values corresponding to the two intervention conditions in each intervention combination, wherein the intervention combination is obtained by traversing the combination of two different intervention conditions among the multiple intervention conditions.
[0095] In some embodiments, the samples are divided into a plurality of groups corresponding to a plurality of intervention conditions, and the corresponding intervention condition is applied to the samples in each group.
[0096] In some embodiments, the construction module 5100 is further configured to calculate the information gain of each current candidate feature by utilizing the difference in feedback information of the samples corresponding to the node to be processed under multiple intervention conditions, wherein the node to be processed is the current leaf node of the decision tree, and the current candidate feature is a feature that has not yet been used to divide the nodes in the decision tree; based on the information gain of each candidate feature, select the candidate feature for division; and use the alternative feature for division to divide the node to be processed and its corresponding samples to generate child nodes of the node to be processed and determine the samples corresponding to each child node.
[0097] In some embodiments, the construction module 5100 is further configured to, for each alternative feature: use the alternative feature to pre-divide the samples corresponding to the node to be processed, and generate a first sample set and a second sample set corresponding to the alternative feature; determine the difference in feedback information between the first sample set and the second sample set under multiple intervention conditions as the information gain of the alternative feature.
[0098] In some embodiments, the construction module 5100 is further configured to, for each set in the first sample set and the second sample set corresponding to the alternative feature, traverse the combination of two intervention conditions in multiple intervention conditions, calculate the difference between the mean values of the feedback values of the samples in the set under the two intervention conditions in each combination, and calculate the sum of these differences as the feedback information of the set; determine the information gain of the alternative feature based on the difference in feedback information between the first sample set and the second sample set.
[0099] In some embodiments, the number of samples corresponding to the node to be processed is greater than the sample threshold.
[0100] In some embodiments, the construction module 5100 is further configured to complete the construction of the decision tree when the depth of the decision tree reaches a depth threshold.
[0101] In some embodiments, the determination module 5200 is further configured to determine, for each intervention condition, the mean of the feedback values of the samples corresponding to the leaf node under the intervention condition as the expected value under the intervention condition corresponding to the leaf node; and determine the prediction result corresponding to the leaf node based on the expected value corresponding to the leaf node under each intervention condition.
[0102] In some embodiments, the determination module 5200 is further configured to determine the expected value corresponding to the leaf node under each intervention condition as the prediction result corresponding to the leaf node.
[0103] In some embodiments, the determination module 5200 is further configured to compare, for each intervention condition, the expected value under the intervention condition corresponding to the leaf node with the corresponding intervention result threshold to determine the effectiveness information of the intervention condition; and determine the effectiveness information of each intervention condition as the predicted result corresponding to the leaf node.
[0104] In some embodiments, the plurality of intervention conditions comprises more than 3 intervention conditions.
[0105] Reference below Figure 6 Embodiments of the disclosed intervention outcome prediction system are described.
[0106] Figure 6 FIG. 4 is a schematic diagram showing a structure of a prediction system for intervention results according to some embodiments of the present disclosure. Figure 6 As shown, the intervention result prediction system 60 of this embodiment includes: a generation device 500 of any one of the aforementioned intervention result prediction models; and a prediction device 610, which is configured to input the sample to be tested into a decision tree, wherein the decision tree is generated by the generation device of the intervention result prediction model; determine the leaf node to which the sample to be tested belongs in the decision tree; and determine the intervention result of the sample to be tested under multiple intervention conditions based on the prediction result corresponding to the leaf node.
[0107] Figure 7 FIG. 2 shows a schematic diagram of a structure of a device for generating an intervention result prediction model according to other embodiments of the present disclosure. Figure 7 As shown, the device 70 for generating the intervention outcome prediction model of this embodiment includes: a memory 710 and a processor 720 coupled to the memory 710, and the processor 720 is configured to execute the method for generating the intervention outcome prediction model in any one of the aforementioned embodiments based on instructions stored in the memory 710.
[0108] The memory 710 may include, for example, a system memory, a fixed non-volatile storage medium, etc. The system memory may store, for example, an operating system, an application program, a boot loader, and other programs.
[0109] Figure 8 FIG. 2 shows a schematic diagram of a structure of a device for generating an intervention result prediction model according to some other embodiments of the present disclosure. Figure 8 As shown, the generation device 80 of the intervention result prediction model of this embodiment includes: a memory 810 and a processor 820, and may also include an input and output interface 830, a network interface 840, a storage interface 850, etc. These interfaces 830, 840, 850 and the memory 810 and the processor 820 can be connected, for example, through a bus 860. Among them, the input and output interface 830 provides a connection interface for input and output devices such as a display, a mouse, a keyboard, and a touch screen. The network interface 840 provides a connection interface for various networked devices. The storage interface 850 provides a connection interface for external storage devices such as SD cards and USB flash drives.
[0110] An embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements any of the aforementioned methods for generating an intervention outcome prediction model.
[0111] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable non-transient storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0112] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0113] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0115] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure should be included in the protection scope of the present disclosure.
Claims
1. A method for generating an intervention outcome prediction model, comprising: The sample and the difference in feedback information of the sample under multiple intervention conditions are used as information gain of the sample's features, so as to construct a decision tree using the information gain, wherein non-leaf nodes in the decision tree are divided using the sample's features to generate child nodes; For each leaf node of the decision tree, the prediction result corresponding to the leaf node is determined according to the feedback value of the sample corresponding to the leaf node under the multiple intervention conditions, so that the decision tree can be used to predict the intervention results of the multiple intervention conditions.
2. The generation method according to claim 1, wherein: The information gain is determined in the process of constructing the decision tree according to the difference in feedback information of different sample sets under multiple intervention conditions, and the different sample sets are generated by pre-dividing using the characteristics of the samples.
3. The generation method according to claim 2, wherein: The feedback information of each sample set is: when multiple intervention conditions are applied to the samples in the sample set respectively and the mean of the sample feedback value corresponding to each intervention condition is determined, the sum of the differences between the means of the feedback values corresponding to the two intervention conditions in each intervention combination, wherein the intervention combination is obtained by traversing the combination of two different intervention conditions among the multiple intervention conditions.
4. The generation method according to claim 2, wherein: The samples are divided into a plurality of groups corresponding to the plurality of intervention conditions, and the corresponding intervention condition is applied to the samples in each group.
5. The data processing method according to any one of claims 1 to 4, wherein: The using the sample and the difference in feedback information of the sample under multiple intervention conditions as the information gain of the sample's features to construct a decision tree using the information gain includes: Calculate the information gain of each current candidate feature by using the difference in feedback information of the samples corresponding to the node to be processed under the multiple intervention conditions, wherein the node to be processed is the current leaf node of the decision tree, and the current candidate feature is a feature that has not been used to divide the nodes in the decision tree; According to the information gain of each candidate feature, select the candidate feature for partitioning; The node to be processed and its corresponding samples are divided using the candidate features for division to generate child nodes of the node to be processed and determine the samples corresponding to each child node.
6. The generation method according to claim 5, wherein: The calculating the information gain of each current candidate feature by using the difference of feedback information of the samples corresponding to the node to be processed under the multiple intervention conditions comprises: For each candidate feature: Pre-dividing the samples corresponding to the to-be-processed nodes by using the candidate features to generate a first sample set and a second sample set corresponding to the candidate features; The difference in feedback information between the first sample set and the second sample set under the multiple intervention conditions is determined as the information gain of the candidate feature.
7. The generation method according to claim 6, wherein: The determining of the difference in feedback information between the first sample set and the second sample set under the multiple intervention conditions as the information gain of the candidate feature includes: For each of the first sample set and the second sample set corresponding to the candidate feature, traverse the combination of two intervention conditions in the multiple intervention conditions, calculate the difference between the mean values of the feedback values of the samples in the set under the two intervention conditions in each combination, and calculate the sum of these differences as the feedback information of the set; The information gain of the candidate feature is determined according to the difference between the feedback information of the first sample set and the second sample set.
8. The generation method according to claim 5, wherein: The number of samples corresponding to the node to be processed is greater than the sample threshold.
9. The generation method according to any one of claims 1 to 4, wherein: When the depth of the decision tree reaches a depth threshold, the construction of the decision tree is completed.
10. The generation method according to any one of claims 1 to 4, wherein: Determining the prediction result corresponding to the leaf node according to the feedback value of the sample corresponding to the leaf node under the multiple intervention conditions includes: For each intervention condition, the mean of the feedback values of the samples corresponding to the leaf node under the intervention condition is determined as the expected value under the intervention condition corresponding to the leaf node; According to the expected value corresponding to the leaf node under each intervention condition, the prediction result corresponding to the leaf node is determined.
11. The generation method according to claim 10, wherein: Determining the prediction result corresponding to the leaf node according to the expected value corresponding to the leaf node under each intervention condition includes: The expected value corresponding to the leaf node under each intervention condition is determined as the prediction result corresponding to the leaf node.
12. The generation method according to claim 10, wherein: Determining the prediction result corresponding to the leaf node according to the expected value corresponding to the leaf node under each intervention condition includes: For each intervention condition, comparing the expected value under the intervention condition corresponding to the leaf node with the corresponding intervention result threshold to determine the validity information of the intervention condition; The validity information of each intervention condition is determined as the prediction result corresponding to the leaf node.
13. The generation method according to any one of claims 1 to 4, wherein: The multiple intervention conditions include more than 3 intervention conditions.
14. The generation method according to any one of claims 1 to 4, wherein: The sample is a biological sample, the intervention condition includes at least one of a drug type or a drug dosage, and the feedback information represents a response of the biological sample after the intervention condition is applied; or The sample is a user of the application, the intervention condition includes at least one of resources or information delivered to the user, and the feedback information represents the interaction operation of the user with the application after the intervention condition is applied.
15. A method for predicting intervention results, comprising: Inputting the sample to be tested into a decision tree, wherein the decision tree is generated by using the method for generating an intervention outcome prediction model according to any one of claims 1 to 14; Determine the leaf node to which the sample to be tested belongs in the decision tree; According to the prediction results corresponding to the leaf nodes, the intervention results of the sample to be tested under multiple intervention conditions are determined.
16. A device for generating an intervention outcome prediction model, comprising: A construction module is configured to use the sample and the difference in feedback information of the sample under multiple intervention conditions as the information gain of the sample's characteristics to construct a decision tree, wherein non-leaf nodes in the decision tree are divided using the sample's characteristics to generate child nodes; The determination module is configured to determine, for each leaf node of the decision tree, a prediction result corresponding to the leaf node according to the feedback value of the sample corresponding to the leaf node under the multiple intervention conditions.
17. A device for generating an intervention outcome prediction model, comprising: Memory; as well as A processor coupled to the memory, wherein the processor is configured to execute the method for generating an intervention outcome prediction model according to any one of claims 1 to 14 based on instructions stored in the memory.
18. A system for predicting intervention results, comprising: A device for generating an intervention outcome prediction model as claimed in claim 16 or 17; as well as A prediction device, configured to input the sample to be tested into a decision tree, wherein the decision tree is generated by the generation device of the intervention result prediction model; Determine the leaf node to which the sample to be tested belongs in the decision tree; and determine the intervention result of the sample to be tested under multiple intervention conditions according to the prediction result corresponding to the leaf node.
19. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the method for generating an intervention outcome prediction model according to any one of claims 1 to 14 is implemented.