Decision tree growing method, system and equipment based on association rules

Abstract

The invention belongs to the field of command and control system construction decision, and particularly relates to a decision tree growing method, system and device based on association rules. Extracting mining features based on the original data sample; based on the mining features, discretizing data in the original data sample to obtain transaction items corresponding to different mining features, and forming to-be-mined transactions by different transaction items; performing association rule mining on the to-be-mined transaction to obtain an association rule between transaction items; obtaining an initial decision tree of the mining features; and fusing the association rule into the initial decision tree to obtain an optimized decision tree. The method has the effects of reducing interference of invalid information and quickly constructing the decision tree.

Description

Technical Field

[0001] The present invention belongs to the field of decision-making in charge and control system construction, and in particular relates to a decision tree growth method, system and equipment based on association rules. Background Art

[0002] In the future multi-domain combat and intelligent combat scenarios, the resources available to the command and control system have the characteristics of wide distribution, multiple sources, multiple types, complex relationships and cross-domain. The substantial expansion of types and time and space scope has increased the scale and complexity of system construction exponentially. When facing changes in combat organization, command relations and command objects, the average reconstruction time of the current command and control system is several hours to several days, which is difficult to adapt to the fast combat rhythm in the future. With the development of software-defined technology and the continuous deepening of the informatization of our military equipment, the proportion of digital and informationized parts in equipment is increasing, and the application functions are increasing. The software-defined command and control system has become one of the important development directions of the command information system architecture in the future. The software-defined command and control system requires the system to be able to autonomously perceive environmental changes, autonomously adjust the architecture plan, realize human-in-the-loop combat control, and continuously acquire and learn dynamic knowledge to enable the system to adapt to the continuous growth of combat capabilities. During the operation of the software-defined command and control system, due to changes in the combat environment such as combat mission adjustments, system node failures, or network communication environment degradation, the original organizational structure, business processes, task allocation methods, and command and control links of the command and control system are no longer adapted to the new combat environment. Therefore, the system needs to make adaptive adjustment decisions autonomously and promptly, generate adjustment plans, and implement structural or service adjustments.

[0003] In the related technologies, the adaptive decision-making of the control system is traditionally based on strategy or model. The strategy-based decision-making method refers to making decisions based on pre-set strategies. The strategy describes the actions to be taken under given circumstances and conditions. The decision-making process is based on the selection and execution of strategies to achieve the desired goals; the model-based decision-making method refers to making decisions based on established mathematical or logical models. The model can be a statistical model, a mathematical model, a simulation model, etc. By analyzing and predicting the model, the optimal decision-making plan can be obtained.

[0004] Regarding the above-mentioned related technologies, whether it is a policy-based or model-based decision-making method, a large amount of invalid information is included in the decision-making process, and the invalid information will affect the construction and performance of the decision tree. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a decision tree growth method, system and equipment based on association rules, and to construct a decision tree by mining association rules, thereby reducing the interference of invalid information irrelevant to radar blind spot filling or information with low correlation, so that the prediction results through the decision tree are more accurate.

[0006] A decision tree growing method based on association rules, comprising:

[0007] Get the original data sample;

[0008] Extracting mining features based on the original data sample;

[0009] Based on the mining features, discretize the data in the original data sample to obtain transaction items corresponding to different mining features, and different transaction items constitute transactions to be mined;

[0010] Performing association rule mining on the transactions to be mined to obtain association rules between transaction items;

[0011] Obtaining an initial decision tree of the mining features;

[0012] The association rules are integrated into the initial decision tree to obtain an optimized decision tree.

[0013] The original data samples are relevant data on insufficient radar detection area coverage, including data sequence number, blind spot quantity data, warning level data and blind spot area data. The mined features are features of samples in the original samples, including blind spot quantity, blind spot area and warning level.

[0014] Association rules are used to mine the relationship between features. When one of the mined features is used as a condition, the probability of another mined feature also exists. That is, under a certain number of blind spots, a certain blind spot area may also appear, under a certain blind spot area, a certain alarm level may also appear, etc. The transaction item is a collection of data items mined through association rules. The initial decision tree uses the number of blind spots, blind spot area and alarm level as opportunity nodes, and specific solutions as strategy selection nodes. The solution is the number and area of ​​blind spots for each radar.

[0015] Specifically, in decision theory, a decision tree is a prediction model that represents a mapping relationship between object attributes and object values. Each node in the tree represents an object, each forked path represents a possible attribute value, and each leaf node corresponds to the value of the object represented by the path from the root node to the leaf node.

[0016] The decision tree is a hierarchical tree structure consisting of root nodes, branches, internal nodes, and leaf nodes. The decision tree contains three types of nodes: solution decision nodes, chance nodes, and strategy selection nodes.

[0017] The root node, i.e., the solution decision node, can make decisions according to certain problem characteristics and a certain problem decomposition method, and represents the solution decision for the problem. For example, for the problem of insufficient radar detection area coverage, the solution decision node can be (which) combinations all areas should be divided into and (which) areas should be included in each combination.

[0018] Internal nodes are opportunity decision nodes, which judge the decision process according to the data attributes related to the decision problem and represent the path of the decision process. For example, for the problem of insufficient coverage of radar detection area, the opportunity decision node can be the area of ​​radar blind spots, the number of radar blind spots, the number of available radars, etc.

[0019] The leaf node is the strategy selection node, and the strategy can be determined according to certain problem characteristics. The leaf node of the decision tree represents the meta-strategy rather than the final implementation strategy.

[0020] The growth of a decision tree refers to the process of constructing a decision tree by recursively partitioning a data set. In special circumstances, the system can be quickly reconstructed to generate decision strategies and achieve autonomous adaptive decision-making capabilities.

[0021] Specifically, a decision tree growing method based on association rules also includes:

[0022] Acquire raw data samples of insufficient radar detection area coverage, wherein the raw data samples include data sequence number, blind spot quantity data, warning level data, and blind spot area data;

[0023] Extracting mining features based on the original data samples, the mining features including the number of blind spots, the area of ​​the blind spots, and the alarm level;

[0024] Based on the mining features, discretize the data in the original data sample to obtain transaction items corresponding to different mining features, and different transaction items constitute transactions to be mined;

[0025] Performing association rule mining on the transactions to be mined to obtain association rules between transaction items, wherein the association rules are the probabilities of the remaining mining features appearing when one of the mining features is used as a condition;

[0026] Obtain an initial decision tree of the mining feature, wherein the initial decision tree includes the hierarchical order of the mining feature in the decision tree; integrate the association rule into the initial decision tree to obtain an optimized decision tree

[0027] Optionally, based on the mining features, discretizing the data in the original data sample to obtain transaction items corresponding to different mining features includes:

[0028] Initializing K cluster centers of the original data sample;

[0029] Calculate the spatial distance between each data point in the original data sample and the K cluster centers, and assign each data point to the cluster with the nearest cluster center;

[0030] Calculate the average distance from all data points in a cluster to the cluster center, and use the average as the new cluster center;

[0031] Determine whether the new cluster center is the same as the cluster center;

[0032] If the new cluster center is the same as the original cluster center, a clustering result of each mining feature is obtained;

[0033] Based on the original data sample, calculating the silhouette coefficient;

[0034] Changing the K value based on the silhouette coefficient to obtain an optimal silhouette coefficient;

[0035] An optimal clustering result corresponding to the optimal silhouette coefficient is obtained, and the optimal clustering result is used as a transaction item corresponding to different mining features.

[0036] Optionally, the replacing the K value based on the silhouette coefficient to obtain an optimal silhouette coefficient includes:

[0037] Calculate the silhouette coefficient under different K values;

[0038] The silhouette coefficients under different K values ​​are compared to obtain a maximum silhouette coefficient, and the maximum silhouette coefficient is used as the optimal silhouette coefficient.

[0039] Optionally, performing association rule mining on the to-be-mined transaction to obtain association rules between transaction items includes:

[0040] Calculate the support of each transaction item;

[0041] Scan the transaction to be mined to obtain a candidate item set;

[0042] Based on the candidate item set and the support, a frequent item set is obtained;

[0043] The interest degrees between different transaction items in the frequent item set are calculated to obtain association rules between the transaction items.

[0044] Optionally, integrating the association rule into the initial decision tree to obtain an optimized decision tree includes:

[0045] Obtaining the mining feature level of the initial decision tree;

[0046] Obtaining the mining feature level and value range corresponding to the mined feature of the association rule;

[0047] Based on the mining feature level and value range, the association rules are integrated into the initial decision tree to obtain an optimized decision tree.

[0048] Optionally, the step of integrating the association rules into the initial decision tree based on the mining feature level and value range to obtain an optimized decision tree includes:

[0049] Matching the mined feature level with the mining feature level to obtain a merged node of the mining level;

[0050] Based on the integration node and the value range, the association rule is integrated into the initial decision tree to obtain an optimized decision tree.

[0051] Optionally, before discretizing the data in the original data sample based on the mining feature, the method includes:

[0052] Obtaining data types corresponding to different mining features;

[0053] Determine whether the data type is discrete data;

[0054] If the data type is discrete data, the data in the original data sample is discretized based on the mining feature.

[0055] A decision tree growing system based on association rules, comprising:

[0056] An acquisition module, used to obtain raw data samples;

[0057] An extraction module, used for extracting mining features based on the original data sample;

[0058] A discretization module is used to discretize the data in the original data sample based on the mining features to obtain transaction items corresponding to different mining features, and different transaction items constitute transactions to be mined;

[0059] A mining module, used to mine association rules for the transactions to be mined, and obtain association rules between transaction items;

[0060] A second acquisition module is used to acquire an initial decision tree of the mining feature;

[0061] The construction module is used to integrate the association rules into the initial decision tree to obtain an optimized decision tree.

[0062] A terminal device comprises a memory and a processor. The memory stores a computer program that can be run on the processor. When the processor loads and executes the computer program, a decision tree growth method based on association rules is adopted.

[0063] A computer-readable storage medium stores a computer program. When the computer program is loaded and executed by a processor, a decision tree growing method based on association rules is adopted.

[0064] The beneficial effect of the invention is: by mining the original data sample, i.e., the data related to the insufficient coverage of the radar detection area, mining features, i.e., the number of blind spots, the area of ​​blind spots, and the alarm level are obtained, and then the continuous data in the mining features are discretized to obtain transaction items, and then the mining transactions are mined by association rules, and the mined association rules are merged with the initial decision tree to obtain an optimized decision tree, so that on the basis of the initial decision tree, the probability of another mining feature appearing when one of the features is used as a condition can be predicted, and the growth of the decision tree is realized, which provides a solution to the insufficient coverage of the radar detection area, and allocates the number of blind spots and the area of ​​blind spots for each radar. By converting the decision tree growth problem into an association rule mining and analysis problem, the mining and growth of the association rules of the system adaptive decision tree in the big data environment are realized. Compared with the traditional system adaptive decision tree, since the mined association rules are all related, the interference of invalid information can be reduced, making the prediction results through the decision tree more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A flowchart of a decision tree growing method based on association rules of the present invention;

[0066] Figure 2 This is a schematic diagram of discretization of original data samples of the present invention;

[0067] Figure 3 It is a schematic diagram of the decision tree of the autonomous adaptive strategy of the present invention;

[0068] Figure 4 This is a schematic diagram of the initial decision tree of the present invention;

[0069] Figure 5 Schematic diagram of decision tree optimization for the present invention. DETAILED DESCRIPTION

[0070] A decision tree growing method based on association rules, such as Figure 1 As shown, including:

[0071] S100: Obtain original data samples.

[0072] Specifically, the original data sample includes the data sequence number and different mining features, which are usually presented in the form of a table. In this embodiment, taking the problem of insufficient coverage of the radar detection area as an example, the original data sample includes the number of blind spots, the area of ​​the blind spots, and the alarm level, which are usually presented in the form of a table, such as Figure 2 shown.

[0073] S110, extracting mining features based on the original data samples.

[0074] Specifically, the mining features are features or attributes of the samples, such as the number of blind spots, the area of ​​blind spots, and the alarm level. The number of blind spots, the area of ​​blind spots, and the alarm level are the mining features.

[0075] S120. Based on the mining features, discretize the data in the original data sample to obtain transaction items corresponding to different mining features. Different transaction items constitute the transactions to be mined.

[0076] Specifically, different mining features correspond to different types of data. The blind spot quantity data item is a discrete value with a value range of [1,9], the total blind spot area data item is a continuous value with a value range of (0,1000], the blind spot area average data item is a continuous value, and the alarm level data item is a discrete value with a value range of [1,5]. Converting continuous data into discrete data can facilitate association rule mining and improve the effect of association rule mining.

[0077] S130 , performing association rule mining on the transaction to be mined to obtain association rules between transaction items.

[0078] Specifically, association rules are patterns that discover correlations and dependencies between variables in data. They are used to describe the associations between different items in a given data set, such as the association between the number of blind spots and the area of ​​the blind spots. Association rule mining aims to discover frequent item sets and association rules in data to determine the relationships and dependencies between these items.

[0079] S140: Obtain an initial decision tree for mining features.

[0080] Specifically, the initial decision tree includes the hierarchical order of mining features. The decision tree is displayed in the form of branching branches, which represents a mapping relationship between object attributes and object values. Each node in the tree represents an object, and each branching path represents a possible attribute value, and each leaf node corresponds to the value of the object represented by the path from the root node to the leaf node.

[0081] like Figure 3As shown, for example, in the problem of insufficient coverage of the radar detection area, the root node is the mathematical representation of the radar blind spot problem (including the number of blind spots, the location of blind spots, and the size of blind spots, etc.), the first node is the number of radar blind spots, the second node is the area of ​​the radar blind spots, and the third node is the size of the radar blind spots, randomly selecting areas, selecting radars nearby, and maximizing the coverage factor.

[0082] S150, integrating the association rules into the initial decision tree to obtain an optimized decision tree.

[0083] Specifically, when the mined association rules are integrated into the decision tree, we first need to find the nodes corresponding to the mined rules (the position of the same association rules in the decision tree), and then the value range of the mined rules. These two are integrated into the decision tree as conditions to achieve automatic growth of the decision tree.

[0084] In one implementation of this embodiment, step S120 discretizes the data in the original data sample based on the mining features to obtain transaction items corresponding to different mining features, and constructs transactions to be mined according to different transaction items, including:

[0085] S200: Initialize K cluster centers of the original data samples.

[0086] Specifically, the K-Means algorithm is used to perform cluster analysis on each column of data. First, K points in the feature space are randomly set as initial cluster centers.

[0087] S210, calculating the spatial distance from each data point in the original data sample to the K cluster centers, and assigning each data point to the cluster where the nearest cluster center is located.

[0088] S220, calculating the average distance from all data points in a cluster to the cluster center, and taking the average value as the new cluster center.

[0089] Specifically, for each other point, the distance to the K centers is calculated, and the nearest cluster center point is selected as the marked category for the unknown point. Then, based on the marked cluster center, the new center point of each cluster, that is, the new cluster center (average value) is recalculated; if the calculated new center point is the same as the original center point, then the process ends, otherwise, step S210 is repeated.

[0090] S230: Determine whether the new cluster center is the same as the cluster center.

[0091] S240: If the new cluster center is the same as the original cluster center, a clustering result of each mined feature is obtained.

[0092] Specifically, when the center of the new cluster is the same as the center of the original cluster, it means that the best clustering result has been obtained under the current K value.

[0093] S250: Calculate the silhouette coefficient based on the original data sample.

[0094] Specifically, to determine the optimal value of clustering K, the silhouette coefficient is used to evaluate the clustering effect. The calculation method is to calculate the difference between the distance between the sample and other samples in its cluster and the distance between the sample and other clusters outside its cluster, and then divide the difference by the larger value of the two. The closer this value is to 1, the better the clustering effect is; the closer this value is to -1, the worse the clustering effect is; the closer this value is to 0, the clustering effect is not obvious. Assume that the number of clusters is k, and the clusters are p respectively. 1 ,p 2 ,…,p k , sample i belongs to cluster p i , then the silhouette coefficient of sample i is defined as:

[0095]

[0096] Among them, b i represents the cohesion of sample i, dist(i,j) is the distance between sample i and sample j, i and j are samples in the same cluster, max{c i ,b i}To find c i and b i The maximum value of .

[0097]

[0098]

[0099] Among them, c i Represents the separation degree of sample i, and i and q are samples in different clusters.

[0100] The silhouette coefficient of the overall clustering is:

[0101]

[0102] S260: Change the K value based on the silhouette coefficient to obtain the optimal silhouette coefficient.

[0103] Changing the K value based on the silhouette coefficient to obtain the optimal silhouette coefficient includes:

[0104] S2600, calculating the silhouette coefficient under different K values.

[0105] S2610. Compare the silhouette coefficients under different K values, obtain the maximum silhouette coefficient, and use the maximum silhouette coefficient as the optimal silhouette coefficient.

[0106] Specifically, the silhouette coefficients under different K values ​​are calculated, and the largest silhouette coefficient is found. The change in the K value is determined by the step size. The K value corresponding to the largest silhouette coefficient is the required value. Clustering is performed under the current K value to obtain the optimal clustering result. The optimal clustering result is composed of different item sets (transaction items), and multiple item sets constitute mining transactions.

[0107] S270: Obtain an optimal clustering result corresponding to the optimal silhouette coefficient, and use the optimal clustering result as the transaction item corresponding to different mining features.

[0108] In one implementation of this embodiment, step S130 is to perform association rule mining on the transaction to be mined, and the association rules between the transaction items include:

[0109] S300: Calculate the support of each transaction item.

[0110] S310: Scan the transactions to be mined to obtain candidate item sets.

[0111] S320: Based on the candidate item set and the support, a frequent item set is obtained.

[0112] S330: Calculate the interest degrees between different transaction items in the frequent item set to obtain association rules between the transaction items.

[0113] Specifically, support is an indicator used in data mining to measure the frequency or probability of an item set appearing in the overall data set. In association rule mining, support describes the prevalence or commonness of an item set in a data set. It can be used to determine whether an item set has sufficient associations, or as a basis for selecting frequent item sets. Support can be calculated using the following formula: Support = Number of times an item set appears / Size of the overall data set.

[0114] The preprocessed data set D can be represented as a set of n transactions, D = {t 1 ,t i+1 ,…,i n}, where each transaction t i ={tid i ,X i ,Y i}, Y represents the meta-strategy data item, |Y i |=1, X represents other environmental characteristic data items, |X i |≥1, then the association rule of the decision tree is represented as the relationship X→Y. Apriori is a classic association rule mining algorithm used to discover frequent patterns.

[0115] Scan the entire original data sample. The first candidate item set is the set of all scanned items. Scanning is to traverse every transaction item in the original data sample. The set of all different transaction items obtained is the candidate item set. Next, find the frequent (k+1)-item set L (k+1) , first connect the frequent k-item set with itself to generate a candidate (k+1)-item set, denoted as C (k+1) , sort the items in order and connect the items that belong to the same item set. At the same time, generate frequent item sets based on the support.

[0116] L k It is C k So from the concept of subset, we can know that C k The members of may or may not be frequent, and L k All subsets of must be in C k Re-traverse the database, for C k , obtain the number of occurrences and proportions of each item, obtain the support, delete the item sets whose support does not meet the requirements, and thus determine L k . And the candidate item set C k This may involve a large amount of computation. If the (k-1)-subset of the candidate k-itemset is not in the frequent item set L k-1 In, L k-1 The superset of C, that is, the candidate k-item set must not occur frequently, and can be directly k . This cycle repeats, traversing the database, and ends when no new frequent item sets are output.

[0117] There are certain defects in the selection of association rules based on support and confidence. If the support and confidence thresholds are too low, a large number of redundant or meaningless rules will be generated. If the thresholds are set too high, some potential association rules of user concern may be filtered out. Therefore, this paper uses interest to select association rules. The specific calculation method is as follows:

[0118]

[0119] Among them, ial(X→Y) is the interest, supp(Y) represents the support of transaction item Y, and conf(X→Y) represents the confidence of the association rule "X→Y".

[0120] The interest measure satisfies ial(X→Y)∈[-1,1]. If the interest ial(X→Y)=0, it means that the item sets X and Y in the pattern are independent of each other; if ial(X→Y)>0, it means that the item sets X and Y in the pattern are positively correlated; if the interest ial(X→Y)<0, it means that the item sets X and Y in the pattern are negatively correlated.

[0121] The association rule mining method using interest measurement requires that X→Y meets the following conditions:

[0122]

[0123] sp 0 >0,cf 0 >0,il 0 >0

[0124] Among them, sp 0 and cf 0 It is a hyperparameter, which can be set by supp(X∪Y) as the support when X and Y appear at the same time.

[0125] In one implementation of this embodiment, step S150 is to integrate the association rules into the initial decision tree, and the optimized decision tree includes:

[0126] S400: Obtain mining feature levels of an initial decision tree.

[0127] S410: Obtain the mining feature level and value range corresponding to the mined feature of the association rule.

[0128] S420: Based on the mining feature level and value range, the association rules are integrated into the initial decision tree to obtain an optimized decision tree.

[0129] S4200: Match the mined feature level with the mining feature level to obtain an integration node of the mining level.

[0130] S4201. Based on the integration nodes and value ranges, the association rules are integrated into the initial decision tree to obtain an optimized decision tree.

[0131] Specifically, the mining feature level of the initial decision tree is the order of the mining feature level in the initial decision tree, such as Figure 4 shown.

[0132] When an association rule X→Y is mined, the rule should be reflected in the system adaptive decision tree. Suppose that an association rule X→Y is mined, X={i 1 ,i 2 ,…,i t} represents t feature data items, Y = y s represents a specific meta-strategy s, the order of data items in X conforms to the data attribute decision order in the decision tree (mining feature hierarchy), and data item i j Indicates that the attribute data value is in a specific range, that is (Value range).

[0133] The mining feature level of the mined associated features and the mining feature level of the initial decision tree are integrated into the position of the same level, that is, into the node, and the position where it should be integrated in the same level is found according to its value range.

[0134] Taking the problem of insufficient radar detection area coverage as an example, the initial decision tree mining features from top to bottom are the number of blind spots, blind spot area, alarm level and meta-strategy number. After mining association rules, the optimized decision tree is as follows: Figure 5 As shown in the figure, the red dotted line is the mined association rule.

[0135] A decision tree growing system based on association rules, comprising:

[0136] An acquisition module, used to obtain raw data samples;

[0137] The extraction module is used to extract mining features based on the original data samples;

[0138] The discretization module is used to discretize the data in the original data sample based on the mining features, obtain the transaction items corresponding to different mining features, and form the transactions to be mined according to different transaction items;

[0139] A mining module is used to perform association rule mining on the transaction to be mined and obtain association rules between transaction items;

[0140] The second acquisition module is used to obtain an initial decision tree for mining features;

[0141] The building module is used to integrate the association rules into the initial decision tree to obtain the optimized decision tree.

[0142] An embodiment of the present application also discloses a terminal device, including a memory and a processor. The memory stores a computer program that can be run on the processor. When the processor loads and executes the computer program, a decision tree growth method based on association rules is adopted.

[0143] The terminal device may be a computer device such as a desktop computer, a laptop computer or a cloud server, and the terminal device includes but is not limited to a processor and a memory. For example, the terminal device may also include input and output devices, a network access device and a bus.

[0144] Among them, the processor can adopt a central processing unit (CPU). Of course, according to actual usage, other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc., and this application does not impose any restrictions on this.

[0145] Among them, the memory can be an internal storage unit of the terminal device, such as a hard disk or memory of the terminal device, or it can be an external storage device of the terminal device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD) or a flash memory card (FC) equipped on the terminal device, etc., and the memory can also be a combination of an internal storage unit and an external storage device of the terminal device. The memory is used to store computer programs and other programs and data required by the terminal device. The memory can also be used to temporarily store data that has been output or is to be output, and this application does not impose any restrictions on this.

[0146] Among them, through this terminal device, the decision tree growing method based on association rules in the above embodiment is stored in the memory of the terminal device, and is loaded and executed on the processor of the terminal device for easy use.

[0147] The embodiment of the present application further discloses a computer-readable storage medium, and the computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the decision tree growing method based on association rules in the above embodiment is adopted.

[0148] Among them, the computer program can be stored in a computer-readable medium, the computer program includes computer program code, the computer program code can be in the form of source code, object code, executable file or certain middleware, etc. The computer-readable medium includes any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the computer-readable medium includes but is not limited to the above-mentioned components.

[0149] Among them, through this computer-readable storage medium, the decision tree growth method based on association rules in the above embodiment is stored in a computer-readable storage medium, and is loaded and executed on a processor to facilitate the storage and application of the above method.

[0150] A person skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of protection of the present application is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present application as above, which are not provided in detail for the sake of simplicity.

[0151] One or more embodiments of the present application are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the present application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of the present application should be included in the protection scope of the present application.

Claims

1. A decision tree growing method based on association rules, characterized in that: include: Get the original data sample; Extracting mining features based on the original data sample; Based on the mining features, discretize the data in the original data sample to obtain transaction items corresponding to different mining features, and different transaction items constitute the transactions to be mined; Performing association rule mining on the transactions to be mined to obtain association rules between transaction items; Obtaining an initial decision tree of the mining features; The association rules are integrated into the initial decision tree to obtain an optimized decision tree.

2. The decision tree growing method based on association rules as claimed in claim 1, characterized in that: The original data samples are relevant data on insufficient coverage of radar detection areas, including data serial number, blind spot quantity data, warning level data and blind spot area data. The mined features are features of samples in the original samples, including blind spot quantity, blind spot area and warning level.

3. The decision tree growing method based on association rules as claimed in claim 1, characterized in that: The discretization of the data in the original data sample based on the mining features to obtain transaction items corresponding to different mining features includes: Initializing K cluster centers of the original data sample; Calculate the spatial distance between each data point in the original data sample and the K cluster centers, and assign each data point to the cluster with the nearest cluster center; Calculate the average distance from all data points in a cluster to the cluster center, and use the average as the new cluster center; Determine whether the new cluster center is the same as the cluster center; If the new cluster center is the same as the original cluster center, a clustering result of each mining feature is obtained; Based on the original data sample, calculating the silhouette coefficient; Changing the K value based on the silhouette coefficient to obtain an optimal silhouette coefficient; An optimal clustering result corresponding to the optimal silhouette coefficient is obtained, and the optimal clustering result is used as a transaction item corresponding to different mining features.

4. The decision tree growing method based on association rules as claimed in claim 3, characterized in that: The replacing of the K value based on the silhouette coefficient to obtain the optimal silhouette coefficient comprises: Calculate the silhouette coefficient under different K values; The silhouette coefficients under different K values ​​are compared to obtain a maximum silhouette coefficient, and the maximum silhouette coefficient is used as the optimal silhouette coefficient.

5. The decision tree growing method based on association rules as claimed in claim 1, characterized in that: The performing association rule mining on the to-be-mined transaction to obtain association rules between transaction items comprises: Calculate the support of each transaction item; Scan the transaction to be mined to obtain a candidate item set; Based on the candidate item set and the support, a frequent item set is obtained; The interest levels between different transaction items in the frequent item set are calculated to obtain association rules between the transaction items.

6. The decision tree growing method based on association rules as claimed in claim 1, characterized in that: Integrating the association rules into the initial decision tree to obtain an optimized decision tree comprises: Obtaining a mining feature level of the initial decision tree; Obtaining the mining feature level and value range corresponding to the mining feature of the association rule; Based on the mining feature level and value range corresponding to the mining feature of the association rule, the association rule is integrated into the initial decision tree to obtain an optimized decision tree.

7. The system decision optimization method based on association rules as claimed in claim 6 is characterized in that: The step of integrating the association rules into the initial decision tree based on the mining feature level and value range to obtain an optimized decision tree includes: Matching the mining feature level of the initial decision tree with the mining feature level of the association rule to obtain an integration node of the mining feature level; Based on the integration node and the value range, the association rule is integrated into the initial decision tree to obtain an optimized decision tree.

8. The decision tree growing method based on association rules as claimed in claim 1, characterized in that: The step of discretizing the data in the original data sample based on the mining feature includes: Obtaining data types corresponding to different mining features; Determine whether the data type is discrete data; If the data type is discrete data, the data in the original data sample is discretized based on the mining feature.

9. A decision tree growth system based on association rules, characterized in that: include: An acquisition module, used to obtain raw data samples; An extraction module, used for extracting mining features based on the original data sample; A discretization module is used to discretize the data in the original data sample based on the mining features to obtain transaction items corresponding to different mining features, and different transaction items constitute the transaction to be mined; A mining module, used to mine association rules for the transactions to be mined, and obtain association rules between transaction items; A second acquisition module is used to acquire an initial decision tree of the mining feature; The construction module is used to integrate the association rules into the initial decision tree to obtain an optimized decision tree.

10. A terminal device, comprising a memory and a processor, characterized in that: The memory stores a computer program that can be run on the processor, and when the processor loads and executes the computer program, the growth method according to any one of claims 1 to 8 is adopted.

Images