Localized differential privacy protection frequent item set mining method based on frequent pattern tree

Abstract

The invention discloses a localized differential privacy protection frequent item set mining method based on a frequent pattern tree, which is applied to a scene formed by an untrusted third-party data aggregator A and n users, and comprises the following steps of: S1, an initialization stage; S2, a data pruning stage; S3, a tree building stage; S4, a data mining stage. According to the method, under the condition that the third-party data aggregator A does not hold any user privacy record information, all frequent item sets meeting a given support degree threshold value and corresponding support degrees can be estimated, so that it is guaranteed that a third party can mine useful association rules according to an obtained result, and support is provided for possible decisions.

Description

technical field

[0001] The invention relates to the technical fields of data mining and information security, in particular to a frequent item set mining method based on frequent pattern tree localized differential privacy protection. Background technique

[0002] Frequent Itemset Mining (Frequent Itemset Mining) is usually the most important step of Association Rule Mining (AssociationRule Mining), and it is a very important topic in the research of data mining. Its purpose is to mine the variables that often appear together in the data set, and then provide support possible decisions. Therefore, FIM has a wide range of applications, such as transaction data analysis, website intrusion monitoring, etc. In the current competitive environment of society, in order to achieve mutual benefit during business collaboration, while data is shared among different organizations, people pay more and more attention to the protection of personal privacy information. In order to ensure ...

Examples

Embodiment Construction

[0095] Such as figure 1 As shown, in this embodiment, a localized differential privacy-protected frequent itemset mining method based on frequent pattern trees is applied to an untrustworthy third-party data aggregator A, n users U, and privacy records V. In the scenario, the i-th user u i The privacy record of v i ,and I represents the set of known privacy items, denoted as I={x 1 ,x 2 ,...,x d ,...,x D}, x d Indicates the dth private item, 1≤i≤n, 1≤d≤D, D is the total number of private data; the frequent itemset mining method includes the following steps:

[0096] S1. Initialization phase:

[0097] Assuming that the frequent itemset mining method is applied to a shopping mall to mine the collection of frequently purchased commodities by users, and the user's transaction records are private and unknown to the mall; given the support count threshold σ, the frequent itemsets of n user privacy transaction records Mining, you will get all the commodity sets with occurre...

technical field

[0001] The invention relates to the technical fields of data mining and information security, in particular to a frequent item set mining method based on frequent pattern tree localized differential privacy protection. Background technique

[0002] Frequent Itemset Mining (Frequent Itemset Mining) is usually the most important step of Association Rule Mining (AssociationRule Mining), and it is a very important topic in the research of data mining. Its purpose is to mine the variables that often appear together in the data set, and then provide support possible decisions. Therefore, FIM has a wide range of applications, such as transaction data analysis, website intrusion monitoring, etc. In the current competitive environment of society, in order to achieve mutual benefit during business collaboration, while data is shared among different organizations, people pay more and more attention to the protection of personal privacy information. In order to ensure ...

Examples

Embodiment Construction

[0095] Such as figure 1 As shown, in this embodiment, a localized differential privacy-protected frequent itemset mining method based on frequent pattern trees is applied to an untrustworthy third-party data aggregator A, n users U, and privacy records V. In the scenario, the i-th user u i The privacy record of v i ,and I represents the set of known privacy items, denoted as I={x 1 ,x 2 ,...,x d ,...,x D}, x d Indicates the dth private item, 1≤i≤n, 1≤d≤D, D is the total number of private data; the frequent itemset mining method includes the following steps:

[0096] S1. Initialization phase:

[0097] Assuming that the frequent itemset mining method is applied to a shopping mall to mine the collection of frequently purchased commodities by users, and the user's transaction records are private and unknown to the mall; given the support count threshold σ, the frequent itemsets of n user privacy transaction records Mining, you will get all the commodity sets with occurre...

Images