CN102929896A - Data mining method based on privacy protection - Google Patents

Abstract

The invention belongs to the field of data mining, and particularly relates to a data mining method based on privacy protection. The data mining method based on privacy protection comprises the following steps of: conversion of original real data: discretizing potential digital attribute data, arranging a transition probability matrix for all attributes; generation of a decision tree: obtaining statistics of converted data records, and generating the decision tree gradually and recursively by using a converted training sample data set S, a determined splitting attribute set, splitting points and data subset marks; and the generation of a classification rule: pruning the generated decision tree to generate the classification rule. The data mining method based on privacy protection provided by the invention is applicable to noncharacter type data and nonuniformly distributed original data, and can also convert tag attributes; and moreover a classification tree constructed based on the converted data set has the characteristic of higher accuracy.

Description

Data digging method based on secret protection

Technical field

The invention belongs to Data Mining, relate in particular to the data digging method based on secret protection.

Background technology

Current society is the society of an information explosion, the development of internet has aggravated again exchange and the propagation of information, all these have excited again the demand of excavating useful information from a large amount of data greatly, these data and consequent information are the treasures of each industry, and it is recording the essential situation of managing faithfully.But in the face of like this a large amount of data, traditional data analysing method, can only obtain the surface layer information of data such as data retrieval, statistical study, can not obtain information its inherence, profound, the supvr is faced with the abundant and predicament of knowledge poorness of data.Therefore it is very important how excavating the useful knowledge of business decision from these data.

Summary of the invention

The invention provides a kind of data digging method based on secret protection, it comprises following steps:

The shift step of original True Data is carried out discretize to potential numeric type attribute data, again all properties is arranged transition probability matrix;

The generation step of decision tree, at server end, the data recording of statistics after the conversion utilized training sample data collection S, Split Attribute collection, split point and the data subset sign of having determined, progressively recursive generation decision tree after the conversion;

The generation step of classifying rules is carried out beta pruning to the above-mentioned decision tree that has generated, and produces classifying rules.

The raw data that the data digging method based on secret protection of the present invention invention goes for non-character type data and non-uniform Distribution also can the conversion tag attributes, and the classification tree that the data set after conversion is constructed has higher precision.

Description of drawings

Fig. 1 is the flow chart of steps based on the data digging method of secret protection.

Embodiment

Based on the data digging method of secret protection, its flow chart of steps comprises following steps as shown in Figure 1:

The shift step of original True Data is carried out discretize to potential numeric type attribute data, again all properties is arranged transition probability matrix;

The generation step of decision tree, at server end, the data recording of statistics after the conversion utilized training sample data collection S, Split Attribute collection, split point and the data subset sign of having determined, progressively recursive generation decision tree after the conversion;

The generation step of classifying rules is carried out beta pruning to the above-mentioned decision tree that has generated, and produces classifying rules.

Claims (1)

1. based on the data digging method of secret protection, it is characterized in that, comprise following steps:

The shift step of original True Data is carried out discretize to potential numeric type attribute data, again all properties is arranged transition probability matrix;

The generation step of decision tree, at server end, the data recording of statistics after the conversion utilized training sample data collection S, Split Attribute collection, split point and the data subset sign of having determined, progressively recursive generation decision tree after the conversion;

The generation step of classifying rules is carried out beta pruning to the above-mentioned decision tree that has generated, and produces classifying rules.

Images