ETHISCOUT: AN AI-BASED MULTI-PLATFORM DEVELOPER ETHICS AND BEHAVIOR ANALYSIS SYSTEM USING MACHINE LEARNING

The increasing adoption of online developer platforms such as GitHub and Reddit necessitates robust tools for
evaluating the professional conduct and ethical behaviour of software developers. Traditional methods of manual
review are time-consuming, subjective, and unable to scale to the volume of publicly available data on these
platforms. This paper presents EthiScout v2.0, an AI-Based Multi-Platform Developer Ethics and Behavior
Analysis System that combines VADER sentiment analysis with a machine learning (ML) classification pipeline
to produce a composite Suspicious Score for any candidate. The system is built using Python (Streamlit) for the
web interface, GitHub and Reddit public APIs for data collection, VADER for real-time sentiment scoring, and
Scikit-learn (Logistic Regression and Naive Bayes with TF-IDF vectorisation) for toxicity classification. A
dedicated Training Data module allows users to load Kaggle-compatible datasets, train a model, evaluate its
performance with accuracy metrics and a confusion matrix, and run live text predictions. The composite score is
computed as a weighted aggregate of GitHub (35%), Reddit (35%), and ML Toxicity (30%) sub-scores.
Experimental evaluation on standard toxicity datasets demonstrates classification accuracy of 92.6% with Logistic
Regression and 89.3% with Naive Bayes, with live prediction latency under 10 ms.