A Prediction Model with Multi-Pattern Missing Data Imputation for Medical Dataset

Medical data is over and over again analyzed for disease diagnosis and proper treatment. Medical dataset usually contain missing data it is also treated as error. These missing values possibly will clue to incorrect disease diagnosis result. Meanwhile the medical data collection is costly, time incontrollable and an essential on the way to collected beginning various issues. Therefore get better missing data is an alternative of re-collecting the medical data. In this paper a Prediction Model has been proposed for missing data imputation in medical data. An experiment includes various datasets to validate the model as well as to establish the importance of imputation. A new Method name called enhanced random forest regression predictor is proposed for missing data imputation on medical dataset. Method is validated using 3 datasets named wisconsin, dermatology and breast cancer. All the datasets are downloaded from UCI repository. Missing data are generated manually in the original data from 1% to 15%. The proposed Prediction model is predict the missing values based on enhanced random forest regression predictor and evaluates the model using various classifiers. Classification is assessment of normal and abnormal disease diagnostics and produce the result of this experiment is accuracy. Proposed predictor has been compared with two imputation method as KNN and mice forest. Missing prediction model is perform better compared with other methods. Evaluation is demonstrating the classification and gives accuracy which is compared with original dataset and the imputed dataset. Missing data problem is a serious problem in medical data and can guidance downstream disease analysis. A proposed enhanced missing prediction model for missing data imputation is an application of imputing the missing data and disease analysis using classification in better way.