A random decision tree frameworkfor privacy preserving data mining
Download as DOCX, PDF0 likes57 views
This document summarizes a framework for privacy-preserving data mining using a random decision tree algorithm. Multiple parties like banks, insurance companies, and credit card companies share data but need to keep certain attributes private. The random decision tree algorithm partitions data based on each party's needs, encrypts the data using homomorphic encryption, builds a decision tree model on the encrypted data, and allows parties to classify new instances while preserving privacy. It compares the accuracy of random decision trees to traditional ID3 decision trees.
A random decision tree frameworkfor privacy preserving data mining
- 1. A Random Decision Tree Framework for Privacy-Preserving Data Mining Data mining is used to discover knowledge by using existing or past data and new data class can be find out by applying it on existing using classification technique. Now-a-days multiple parties use same data to identify class name of their data and if we expose all data to all parties then privacy will be at risk. For example multiple parties such as bank, insurance company or credit card company will use same records but for different purposes Bank will use it to find past transaction Credit card will use data attributes related to pass payment Insurance company will use to identify correct policy for that person All above companies will use person profile information but with different attributes. If all data expose to all company then privacy will be at risk. To overcome from such issue author has introduce data mining algorithm called Random Decision Tree which can build tree by randomly selected data and apply homomorphic encryption to provide privacy to users data. All companies only knows class name and dataset will be partition based on the company required. With partition dataset Random decision tree will be build. Dataset will be given to Random decision tree algorithm to build a tree which is also called as classification model. To classify new instance company will give all attributes values related to their provided. Then application will apply new instance (record) on decision tree model to predict or classify class name of that instance. In this paper author has given algorithms such as
- 2. Horizontal Partition: using this algorithm we will partition dataset based on number of parties. Encryption: Using this algorithm we will encrypt data using Homomorphic encryption technique Buildtree: using this algorithm Random decision tree will be build Classify Instance: using this algorithm we will classify new data or record belongs to which class by applying decision tree model. In this paper author has done accuracy comparison between Random Decision Tree and ID3 tree. To implement this algorithms author has used WEKA tool and we are also using same tool java API to develop this project. In this paper author has used MUSHROOM and NURSERY Dataset and we also used same dataset and this dataset is available inside ‘dataset’ folder. All information related to dataset columns you can find inside information folder. Some dataset examples form NURSERY dataset parents,has_nurs,form,children,housing,finance,social,health,clas s usual,proper,complete,1,convenient,convenient,nonprob,recommende d,recommend usual,proper,complete,1,convenient,convenient,nonprob,priority,prior ity All bold words are column names and all below are two records from that dataset and last column contains class name. While uploading new records from test folder those records will not have class name and application will classify and give class name for that new record. See below test values.
- 3. 2.203259994700768E307,1.8832849888521625E307,2.16639771156 39986E306,1.0250756057356276E306,2.2434704351677847E307,2. 2434704351677847E307,3.4845121783368866E306,1.34719204705 2717E307,? 2.203259994700768E307,1.8832849888521625E307,2.16639771156 39986E306,1.0250756057356276E306,2.2434704351677847E307,2. 2434704351677847E307,3.4845121783368866E306,2.06847477167 45147E307,? Above test values are in encrypted format and in last column we can see ? instead of class name as we don’t know it class and application will predict it. Screen shots Double click on ‘run.bat’ file to get below screen In above screen click on ‘Upload Dataset’ button and upload any dataset
- 4. In above screen I am uploading nursery dataset, now click on ‘Open’ button to get below screen Now click on ‘Run Data Partition & Privacy Encryption’ to partition and encrypt data
- 5. In above screen we can see entire dataset records in plain format, if u want to see Homomorphic encrypted data then click on ‘View Encrypted Data’ to get below screen In above screen we can see all records are encrypted and only class name which are in last column are shown to parties. With this encrypted data nobody can understand anything. Now to build tree on this encrypted data click on ‘Run Random Decision Tree’ button to build tree
- 6. In above screen we can see tree generated by random decision and all nodes contains encrypted data and this tree got accuracy as 87%. In last line we can see accuracy. Now click on ‘Build ID3 Tree’ button to generate tree with ID3 technique In above screen we can see ID3 tree also but its accuracy is 71%. Now click on ‘Classify Instance’ button to upload test file and get prediction or classification result. Here if u build decision tree with NURSERY dataset then upload nursery test dataset only
- 7. In above screen I am uploading nursey test dataset and below is classification result In above screen each records contains ‘?’ at last column and in next line application has given or predict it class name. for example in above screen in first record is classified as ‘recommend’.
- 8. Now click on ‘Random Decision & ID3 Tree Accuracy Graph’ button to get below accuracy graph of both algorithms In above graph x-axis represents algorithm name and y-axis represents accuracy of those algorithms. Similarly you can upload MUSHROOM dataset and test