Physical Activity Level Prediction
Large-scale machine learning models to predict physical activity levels from health parameters for lifestyle intervention.
Workflow at a glance
Prepare activity data
Load available health/activity records.
Process features
Transform records for model training.
Compare classifiers
Train and compare classification models.
Review evaluation
Inspect historical course-project results.
Project facts & reported results
Reported result
Historical project result; evaluation not independently reproduced.
Implementation fact
Implementation fact
Project Overview
Problem Statement
Sedentary lifestyles contribute to numerous health issues. Accurate activity level prediction enables personalized interventions. Need for automated classification system handling large-scale health data.
Approach & Methodology
Processed 2.8 million records with features including heart rate, step count, and sleep duration. Implemented Logistic Regression, Decision Tree, and Random Forest with Apache Spark for distributed computing. Applied feature selection using correlation analysis and RFE. Performed hyperparameter tuning with k-fold cross-validation.
Outcome & Scope
Random Forest achieved 78.02% accuracy outperforming Decision Tree (77.48%) and Logistic Regression (69.44%). Feature importance analysis identified heart rate and step count as key predictors. Demonstrated scalability with big data framework. Academic project showcasing ML for preventive healthcare.
Key Features & Capabilities
Big data processing with Spark
Multi-class activity classification
Feature selection and engineering
Hyperparameter optimization
K-fold cross-validation
Confusion matrix and metrics analysis
Current Scope & Limitations
Historical course-project results are not independently reproduced here. Held-out split details, preprocessing boundaries, and metric units are not documented in this case study.