Improving Early-Stage Project Risk Assessment Using Machine Learning and Textual Analysis

Description

Early-stage project evaluation is often based on subjective judgment rather than data-driven analysis. This project aims to improve early-stage project risk assessment by developing a machine learning framework that predicts the likelihood of project success before launch. Building on previous work that used structured features from the Kickstarter dataset, this project will enhance the predictive model through improved feature engineering, hyperparameter optimization, and the integration of textual analysis. Natural Language Processing (NLP) techniques will be applied to analyze project descriptions and extract informative textual features using TF-IDF. Several machine learning models, including Random Forest, XGBoost, LightGBM, and HistGradientBoosting, will be trained and compared. Model performance will be evaluated using standard classification metrics such as Accuracy, Precision, Recall, and F1-score. The project aims to develop a more robust predictive framework and produce a simple risk scoring approach that can support early-stage project evaluation and assist decision-makers in assessing project feasibility before significant resources are committed.