Apriori algorithm-based action rule mining method

Abstract

The invention belongs to the technical field of behavior regulation mining, and particularly relates to an Apriori algorithm-based action rule mining method, which comprises the following steps of: 1, performing data cleaning on situation data, finding out available data, and classifying action rule data into conditions and actions; in the second link, the action rules of each category are called item sets, if the number of times that the various rules appear together in the item sets is larger than the minimum support degree threshold value, the item sets are integrated into frequent item sets, and the evaluation criteria of frequent item set mining comprise the support degree and the confidence degree; according to the method, for data such as situations, a frequent item set library of action class rules is constructed based on an Apriori algorithm, and mining of the action class rules is achieved through frequent item sets.

Description

Technical Field

[0001] This invention belongs to the field of behavior rule mining technology, specifically relating to a method for mining action rules based on the Apriori algorithm. Background Technology

[0002] In various action missions, a large amount of situational data and other types of data have been accumulated. Situational data contains rich scenarios and effective response experience and knowledge. This kind of association rule, which takes the form of "if scenario then response activity", is a kind of empirical knowledge. Therefore, target action rules and response strategies can be extracted from situational data. In the past, the association rule of "if scenario then response activity" could only be used as the basis for action rules. Now, for situational and other data, we can consider mining rules based on the Apriori algorithm. Summary of the Invention

[0003] (a) Technical problems to be solved The technical problem to be solved by this invention is: how to provide an action rule mining method based on the Apriori algorithm.

[0004] (II) Technical Solution To address the aforementioned technical problems, this invention provides a method for mining action rules based on the Apriori algorithm, the method comprising two steps: The first step is to clean the situational data, identify usable data, and classify the action rule data into conditions and actions. The second step involves defining each type of action rule as an itemset. If the number of times all the rules appear together in an itemset exceeds the minimum support threshold, then the itemset is classified as a frequent itemset. The evaluation criteria for frequent itemset mining include support and confidence.

[0005] Here, support represents the joint probability that different items co-occur in the itemset; in a large amount of rule data, if X represents a condition and Y represents an action, for and The support formulas for the two related terms to be analyzed are shown below: Given a frequent itemset, the probability of condition X occurring given that action Y is performed. ; It is the number of times it appears in the frequent itemset. It represents the total number of frequent itemsets.

[0006] Here, confidence level represents the conditional probability of different terms occurring; right The confidence level is: It is the probability that condition X will occur given that action Y is performed. ; It is the probability of performing action Y within a frequent itemset.

[0007] The method utilizes the Apriori algorithm, the principle of which is as follows: enter Data set Support threshold ; Output Top- Frequent itemsets; operate: 1) Traverse the entire dataset to obtain all occurrences of the data, and use them as candidate frequent data points. Itemset; 2) Frequent excavation Itemset: a) Traverse the data to calculate candidates Itemset support; b) Pruning to remove candidate plants The item cluster support is lower than the minimum support. The dataset yields frequent results. Itemsets; if frequent If the itemset is empty, return frequent. The algorithm terminates when the set of itemsets is obtained as the result; otherwise, it terminates. If the itemset contains only one item, return the most frequent item directly. The set of itemsets is the result of the algorithm, and the algorithm ends. c) Based on frequent Itemsets, joined to generate candidate items Itemset; 3) Using step 2), iteratively obtain Itemset results.

[0008] Among them, the Apriori algorithm uses prior knowledge to analyze the problem; If an itemset is a frequent itemset, then its subsets are also frequent itemsets. Utilizing this prior property, the Apriori algorithm can quickly eliminate candidate sets that do not meet the requirements. This type of data describes a relationship between events. Typical association analysis only emphasizes the relationship of simultaneous occurrence. For log data with few itemsets, the Apriori algorithm will not generate too many frequent itemsets, and there is no need to scan the data excessively. When there are few itemsets but many transactions, the superiority of the Apriori algorithm becomes apparent, and its performance can be improved.

[0009] Among them, in the target action category rules, the aircraft reflects the aircraft's true state and decision-making intentions. For air combat decisions, it is necessary to use the aircraft's flight parameters and environmental situation information. The environmental information includes the relative attitudes of the enemy and friendly forces, flight parameters, flight status, and safety status information. In the flight parameters of aircraft, based on the three decision-making criteria of relative attitude, flight parameters and flight status, data preprocessing is carried out on the condition variables. Data quality is improved by means of data cleaning and data transformation. The data is made "clean" by outlier detection and processing, missing value supplementation and other methods. Further abstraction and transformation are then carried out to obtain regular condition variables. Discovering rules requires finding the probability of event X and the probability of event XY occurring, which involves many complex cases and creates a need for searching action rules with multiple conditions and events. Therefore, the original algorithm formula is improved, and the confidence calculation formula is changed as follows: Given that action Y is performed, the probability that condition X and action Y will occur simultaneously is... ; Given a frequent itemset, the probability that condition X and action Y occur simultaneously, given that action Y is performed.

[0010] In the above transformation, the action Y of the action rule is definite and finite in number. From the perspective of action, we first determine an action, and then use (X+Y)-Y to find the rule conditions. The rule generated by the above transformation expression is under certain conditions. This avoids generating rules that ignore them, thereby improving the efficiency of rule discovery.

[0011] (III) Beneficial Effects Compared with existing technologies, this invention constructs a frequent itemset library of action rules based on the Apriori algorithm for situational and other data, and realizes the mining of action rules through frequent itemsets. Attached Figure Description

[0012] Figure 1 , Figure 2This is a schematic diagram illustrating the principle of the technical solution of the present invention. Detailed Implementation

[0013] To make the objectives, contents, and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples.

[0014] To address the aforementioned technical problems, this invention provides a method for mining action rules based on the Apriori algorithm, the method comprising two steps: The first step is to clean the situational data, identify usable data, and classify the action rule data into conditions and actions. The second step involves defining each type of action rule as an itemset. If the number of times all the rules appear together in an itemset exceeds the minimum support threshold, then the itemset is classified as a frequent itemset. The evaluation criteria for frequent itemset mining include support and confidence.

[0015] Here, support represents the joint probability that different items co-occur in the itemset; in a large amount of rule data, if X represents a condition and Y represents an action, for and The support formulas for the two related terms to be analyzed are shown below: Given a frequent itemset, the probability of condition X occurring given that action Y is performed. ; It is the number of times it appears in the frequent itemset. It represents the total number of frequent itemsets.

[0016] Here, confidence level represents the conditional probability of different terms occurring; right The confidence level is: It is the probability that condition X will occur given that action Y is performed. ; It is the probability of performing action Y within a frequent itemset.

[0017] The method utilizes the Apriori algorithm, the principle of which is as follows: enter Data set Support threshold ; Output Top- Frequent itemsets; operate: 1) Traverse the entire dataset to obtain all occurrences of the data, and use them as candidate frequent data points. Itemset; 2) Frequent excavation Itemset: a) Traverse the data to calculate candidates Itemset support; b) Pruning to remove candidate plants The item cluster support is lower than the minimum support. The dataset yields frequent results. Itemsets; if frequent If the itemset is empty, return frequent. The algorithm terminates when the set of itemsets is obtained as the result; otherwise, it terminates. If the itemset contains only one item, return the most frequent item directly. The set of itemsets is the result of the algorithm, and the algorithm ends. c) Based on frequent Itemsets, joined to generate candidate items Itemset; 3) Using step 2), iteratively obtain Itemset results.

[0018] Among them, the Apriori algorithm uses prior knowledge to analyze the problem; If an itemset is a frequent itemset, then its subsets are also frequent itemsets. Utilizing this prior property, the Apriori algorithm can quickly eliminate candidate sets that do not meet the requirements. This type of data describes a relationship between events. Typical association analysis only emphasizes the relationship of simultaneous occurrence. For log data with few itemsets, the Apriori algorithm will not generate too many frequent itemsets, and there is no need to scan the data excessively. When there are few itemsets but many transactions, the superiority of the Apriori algorithm becomes apparent, and its performance can be improved.

[0019] Among them, in the target action category rules, the aircraft reflects the aircraft's true state and decision-making intentions. For air combat decisions, it is necessary to use the aircraft's flight parameters and environmental situation information. The environmental information includes the relative attitudes of the enemy and friendly forces, flight parameters, flight status, and safety status information. In the flight parameters of aircraft, based on the three decision-making criteria of relative attitude, flight parameters and flight status, data preprocessing is carried out on the condition variables. Data quality is improved by means of data cleaning and data transformation. The data is made "clean" by outlier detection and processing, missing value supplementation and other methods. Further abstraction and transformation are then carried out to obtain regular condition variables. Discovering rules requires finding the probability of event X and the probability of event XY occurring, which involves many complex cases and creates a need for searching action rules with multiple conditions and events. Therefore, the original algorithm formula is improved, and the confidence calculation formula is changed as follows: Given that action Y is performed, the probability that condition X and action Y will occur simultaneously is... ; Given a frequent itemset, the probability that condition X and action Y occur simultaneously, given that action Y is performed.

[0020] In the above transformation, the action Y of the action rule is definite and finite in number. From the perspective of action, we first determine an action, and then use (X+Y)-Y to find the rule conditions. The rule generated by the above transformation expression is under certain conditions. This avoids generating rules that ignore them, thereby improving the efficiency of rule discovery.

[0021] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for mining action rules based on the Apriori algorithm, characterized in that, The method comprises two steps: The first step is to clean the situational data, identify usable data, and classify the action rule data into conditions and actions. The second step involves defining each type of action rule as an itemset. If the number of times all the rules appear together in an itemset exceeds the minimum support threshold, then the itemset is classified as a frequent itemset. The evaluation criteria for frequent itemset mining include support and confidence.

2. The action rule mining method based on the Apriori algorithm as described in claim 1, characterized in that, Support represents the joint probability that different items co-occur in an itemset; in a large amount of rule-based data, if X represents a condition and Y represents an action, for and The support formulas for the two related terms to be analyzed are shown below: Given a frequent itemset, the probability of condition X occurring given that action Y is performed. ; It is the number of times it appears in the frequent itemset. It represents the total number of frequent itemsets.

3. The action rule mining method based on the Apriori algorithm as described in claim 2, characterized in that, Confidence level represents the conditional probability of different terms occurring; right The confidence level is: It is the probability that condition X will occur given that action Y is performed. ; It is the probability of performing action Y within a frequent itemset.

4. The action rule mining method based on the Apriori algorithm as described in claim 3, characterized in that, The method described uses the Apriori algorithm, the principle of which is as follows: enter Data set Support threshold ; Output Top- Frequent itemsets; operate: 1) Traverse the entire dataset to obtain all occurrences of the data, and use them as candidate frequent data points. Itemset; 2) Frequent excavation Itemset: a) Traverse the data to calculate candidates Itemset support; b) Pruning to remove candidate plants The item cluster support is lower than the minimum support. The dataset yields frequent results. Itemsets; if frequent If the itemset is empty, return frequent. The algorithm terminates when the set of itemsets is obtained as the result; otherwise, it terminates. If the itemset contains only one item, return the most frequent item directly. The set of itemsets is the result of the algorithm, and the algorithm ends. c) Based on frequent Itemsets, joined to generate candidate items Itemset; 3) Using step 2), iteratively obtain Itemset results.

5. The action rule mining method based on the Apriori algorithm as described in claim 4, characterized in that, The Apriori algorithm uses prior knowledge to analyze problems; If an itemset is a frequent itemset, then its subsets are also frequent itemsets. Utilizing this prior property, the Apriori algorithm can quickly eliminate candidate sets that do not meet the requirements. This type of data describes a relationship between events. Typical association analysis only emphasizes the relationship of simultaneous occurrence. For log data with few itemsets, the Apriori algorithm will not generate too many frequent itemsets, and there is no need to scan the data excessively. When there are few itemsets but many transactions, the superiority of the Apriori algorithm becomes apparent, and its performance can be improved.

6. The action rule mining method based on the Apriori algorithm as described in claim 5, characterized in that, In the target action rules, the aircraft reflects the aircraft's true state and decision-making intentions. Air combat decisions require the aircraft's flight parameters and environmental situation information. Environmental information includes the relative attitudes of the enemy and friendly forces, flight parameters, flight status, and safety status information. In the flight parameters of aircraft, based on the three decision-making criteria of relative attitude, flight parameters and flight status, data preprocessing is carried out on the condition variables. Data quality is improved by means of data cleaning and data transformation. The data is made "clean" by outlier detection and processing, missing value supplementation and other methods. Further abstraction and transformation are carried out to obtain regular condition variables. Discovering rules requires finding the probability of event X and the probability of event XY occurring, which involves many complex cases and creates a need for searching action rules with multiple conditions and events. Therefore, the original algorithm formula is improved, and the confidence calculation formula is changed as follows: Given that action Y is performed, the probability that condition X and action Y will occur simultaneously is... ; Given a frequent itemset, the probability that condition X and action Y occur simultaneously, given that action Y is performed.

7. The action rule mining method based on the Apriori algorithm as described in claim 6, characterized in that, In the above transformation, the action Y of the action rule is definite and finite in number. From the perspective of action, we first determine an action, and then use (X+Y)-Y to find the rule conditions. The rule generated by the above transformation expression is under certain conditions. This avoids generating rules that ignore them, thereby improving the efficiency of rule discovery.

8. The action rule mining method based on the Apriori algorithm as described in claim 1, characterized in that, The method described belongs to the field of behavioral procedure mining technology.

9. The action rule mining method based on the Apriori algorithm as described in claim 1, characterized in that, The method describes a frequent itemset library for action-type rules based on the Apriori algorithm for situational data, and uses these frequent itemsets to mine action-type rules.

10. The action rule mining method based on the Apriori algorithm as described in claim 1, characterized in that, The method described above enables action rule mining using the Apriori algorithm.

Images