3401

3220

كارشناسي ارشد

مهندسي صنايع

اصفهان: دانشگاه صنعتي اصفهان، دانشكده صنايع

1385

دوارده، 166، [II]ص.: مصور، جدول، نمودار

ص. ع به فارسي و انگليسي

محمد رضا زماني

نادر شهاب بوشهري

غلامعلي رئيسي اردلي

مهندسي صنايع و سيستم ها

ID3220

به فارسي و انگليسي:قابل روئت در نسخه ديجيتالي

AbstractFirst an introduction about data mining concepts which are used in thisresearch is provided For this purpose CRISP DM a standard process forimplementing a data mining project is introduced and its steps are described In the next phase these concepts are applied to insurance data The socialsafeguard organization is introduced and some of its needs are describedbriefly Then these needs are formulated to a data mining project Next data preprocessing is performed We evaluate the quality of the Insurancedata clean the raw data deal with missing data and perform transformationson certain variables if needed Next Using charts and graphics such as histograms and scatter plots we try toidentify data set better examine interrelationships among variables as well asdevelop some initial idea of possible associations between predicting variablesand target variable if any Then modeling phase is started Because of the structure of the problem asupervised modeling is applied and using C4 5 algorithm and an applicationnamed WEKA 3 4 four decision trees are created We use two different approaches excluding cost of errors and including cost oferrors for creating the model For either approach two trees two way andmultiple way are developed Then we evaluate trees First using some measures such as True Positive rate False Positive rate True Negative rate False Positive rate Recall Precision F measure MSE Kappa statistics Confusion Matrix and ROC Curves each oftrees are evaluated Second using statistical inference we compare the two wayand multiple way trees Then some classification rules extracted from one of developed trees areprovided Finally we provide some advantages of using models

محمد رضا زماني

نادر شهاب بوشهري

غلامعلي رئيسي اردلي