Smart Stomach Cancer Risk Prediction Application: A Data Mining Approach

Abstract

Stomach Cancer is the 3rd conducting cause of cancer and the 5th most deadly disease among all diseases in world-wide. In this study, our aim is to find out all possible preoperative risk factors of SC and develop an android based application to predict the risk level of SC. From this perspective patient’s data are collected from NICRH. To conduct this study statistical (ANOVA test, ChiSquare test, Odds Ratio, Probability test) and data mining (Feature Selection, Predictive Apriori Algorithm) approach has been used to get significant and highly related risk factors for SC. After that, a risk score algorithm has been designed based on an algorithm and finally developed the application. Experimenting 300 subjects’ records (150 is affected and 150 is non-affected) with 33 risk factors we will get 25 statistically significant P = (P< 0.05) risk factors and 18 top features. Where “Abdominal Pain” is the top preoperative risk factors of SC including (P< 0.000, X2=175.274, and OR = 66.769) and “Nausea” including (P< 0.000, X2=152.261, OR = NA) and “Skin Color Turn into Pale” including (P< 0.000, X2=138.240, and OR =139.462) respectively second and third most risk factors. Also, founded other high-risk factors are “Menetrier Disease = Yes”, “Get Ill Too Much = Yes”, “Previous Stomach Surgery = Yes”, “Take Spicy and Salted Food = Yes”, “Education Level = Less than high school”, “Monthly Income= Less than 20 k”, “Blood Group = A”, “BMI= Severely Underweight or Overweight” and etc. This application will become very helpful and efficient for all researcher, doctors, and peoples from Bangladesh (individually low and middle-income people) to understand the risk factors of SC.