washiemorey/Diabetes-Prediction

Early Diabetes Risk Prediction Tool for Primary Healthcare in Africa # Diabetes-Prediction ### Early Diabetes Risk Prediction Tool for Primary Healthcare in Africa Diabetes is often diagnosed too late in Africa, leading to complications and early deaths. This project aims to build a lightweight, ML-powered tool that predicts the risk of diabetes early using minimal, easily collected clinical data. The goal is to empower primary healthcare workers to triage patients more effectively. Obtained the data from kaggle.Pima Indians Diabetes Dataset Features available: Pregnancies Glucose Blood Pressure Skin Thickness Insulin BMI Diabetes Pedigree Function Age Outcome (0 = No Diabetes, 1 = Diabetes) ### Project Workflow 1. Checked for missing or zero values in glucose, insulin, BMI, BloodPressure, SkinThickness.These are values that can never be 0 in any human being. 2. Replaced 0s with NAs then replaced NAs further with median values. 3. Peformed Exploratory data Analysis on Outcome the see the distribution of diabetes and No Diabetes Cases. 4. Used correlation matrix to identify key features.To find out if there was multicollenearity among any of the indipendent variables. 5. Visualized feature distributions boxplots of Outcome with Glucose and BMI.Very important factors to consider while diagnozing someone for diabetes. ### Model Building (Logistic Regression) Splited the data into two. Training set for training the Model and Testing Set for testing the Model build. After building the model found that some variables like BloodPressure,SkinThickness,Insulin and Age are not significant predictors of Diabetes. Fine tunned the model further by removing the insignificant variables. ### Key Metrics of the Model 1. Accuracy 76.86% implyng that :The model correctly predicts about 77 out of every 100 patients. 2. Precision 61.80% implying that Of all patients predicted to have diabetes, approximately 62% actually do — avoids false alarms. 3. Recall (Sensitivity) 68.75% The model catches nearly 7 out of 10 true diabetic cases — …