RiskPilot: A Web-Based Rule-Based Decision-Support System for Predicting Software Project Success or Failure with Interpretable Risk Indicators

Abstract

Software projects frequently exceed their budgets, miss their schedules or fall short of quality expectations, and manual tracking through spreadsheets rarely exposes these risks early, consistently and with an explanation. This paper presents RiskPilot, a web-based decision-support system that classifies a software project as Success or Failure from twelve project parameters: budget, duration, team size, team experience, project complexity, requirement changes, sprint delays, bug count, testing coverage, resource availability, client involvement and customer satisfaction. The core of the system is a transparent rule-based classifier built from IF–ELSE conditions, which also reports risk indicators and improvement recommendations linked to the rules that were triggered. The prediction service is additionally structured to accept a trained gradient-boosting model with SHAP-based attributions and to fall back to the rule-based classifier when no model is available. The platform provides authenticated access, project management, a risk-distribution dashboard, dataset upload and validation, GitHub repository identification and reporting. The architecture, the prediction-service logic and the implemented modules are described, and the working application is illustrated through screenshots. Quantitative predictive performance on labelled project data has not yet been established; an evaluation protocol is therefore specified. The system is intended to support, not replace, managerial judgement.

Introduction

The RiskPilot – Software Project Success and Failure Prediction System is a web-based application designed to identify potential risks in software development projects and predict whether a project is likely to succeed or fail. Software projects often face challenges such as budget overruns, schedule delays, increasing complexity, frequent requirement changes, insufficient testing, and limited resources. Traditional project management methods often monitor these factors separately, making it difficult to identify risks at an early stage.

The proposed RiskPilot system provides a centralized platform that collects twelve project parameters, evaluates them using predefined IF–ELSE conditions, and classifies the project as either Success or Failure. The system also identifies specific risk factors and provides recommendations to help project managers take corrective actions.

The application includes user registration and login, project creation, parameter entry, prediction history, dashboard visualization, dataset upload, and report generation. It uses a rule-based classification method as its primary prediction approach. When a trained machine-learning model becomes available, the system can also generate model-based predictions with SHAP explanations. If the model fails, the system falls back to the rule-based prediction.

The main purpose of RiskPilot is to make software project risk assessment easier, more transparent, and understandable, even for organizations that do not have large historical datasets.

Main Objectives

1. To predict software project success or failure using predefined IF–ELSE conditions.

2. To identify project risks related to budget, duration, complexity, requirement changes, sprint delays, bugs, and testing.

3. To provide explainable predictions with risk indicators and recommendations.

4. To develop a centralized platform for user authentication, project management, prediction history, and dashboard visualization.

5. To generate project analysis reports and support dataset uploads for future machine-learning model development.

Existing System

Traditional project management systems generally monitor project parameters using spreadsheets or basic dashboards. They often lack integrated predictive analysis and explainable risk identification. Although machine-learning models can predict project outcomes, they require sufficient labelled historical data and proper validation. As a result, organizations without suitable datasets may find it difficult to implement reliable prediction systems.

Proposed System

The proposed RiskPilot system provides an integrated web application that evaluates software project parameters and predicts project outcomes using a transparent rule-based classification approach. It identifies the conditions responsible for potential project failure and provides recommendations to reduce risks.

The system also supports an optional machine-learning prediction method, allowing future integration and comparison of data-driven models with the existing rule-based approach.

Methodology and Implementation

The system follows these major stages:

1. Project Parameter Collection: Collects twelve parameters, including budget, duration, team size, team experience, project complexity, requirement changes, sprint delays, bug count, testing coverage, client involvement, resource availability, and customer satisfaction.

2. Input Validation: Checks whether all required values are entered correctly and fall within the permitted ranges.

3. Rule-Based Classification: Evaluates seven predefined rules to identify project risks.

4. Risk Identification: Displays the triggered risk indicators and highlights the areas requiring attention.

5. Recommendation Generation: Provides suitable suggestions for reducing identified risks.

6. Project Prediction: Classifies the project as Success or Failure based on the number of violated rules.

7. Dashboard and Reporting: Stores prediction results and displays project analysis through dashboards and reports.

8. Machine-Learning Integration: Supports future model training using uploaded datasets and SHAP-based explanations.

Conclusion

This paper presented RiskPilot, a web-based decision-support system that classifies software projects as Success or Failure from twelve parameters using transparent IF–ELSE rules, reports rule-linked risk indicators and recommendations, and is structured to admit a trained model with SHAP explanations. The architecture, prediction-service logic and implemented modules were described and the working interface was illustrated. Because predictive performance has not yet been measured on labelled project data, the paper does not claim a level of accuracy; the usefulness of the system depends on the quality of its rules, the accuracy of the entered data and the outcome of the evaluation set out in Section V-G.

References

[1] T. Kawamura, T. Toma and K. Takano, “Outcome prediction of software projects for information technology vendors,” in Proc. IEEE Int. Conf. Industrial Engineering and Engineering Management (IEEM), Singapore, Dec. 2017, pp. 1733–1737. [2] A. Nizam, “Software project failure process definition,” IEEE Access, vol. 10, pp. 34428–34441, 2022, doi: 10.1109/ACCESS.2022.3162878. [3] R. G. Cooper, “Why AI projects fail: Lessons from new product development,” IEEE Eng. Manag. Rev., vol. 52, no. 4, pp. 15–21, 2024, doi: 10.1109/EMR.2024.3419268. [4] G. D. Taye and Y. A. Feleke, “Prediction of failures in the project management knowledge areas using a machine learning approach for software companies,” SN Appl. Sci. (now Discover Applied Sciences), vol. 4, Art. no. 165, 2022, doi: 10.1007/s42452-022-05051-7. [5] S. M. Lundberg and S.-I. Lee, “A unified approach to interpreting model predictions,” in Advances in Neural Information Processing Systems 30 (NIPS 2017), 2017, pp. 4765–4774; arXiv:1705.07874. [6] T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” in Proc. 22nd ACM SIGKDD Int. Conf. Knowledge Discovery and Data Mining, 2016, pp. 785–794, doi: 10.1145/2939672.2939785.

Copyright

Copyright © 2026 Kaviya R, Latha S, Lavanya S, Mrs. K. Yazhini. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.