Onemalinga/Phase3Project

Tanzania Water Project A PREDICTIVE MODEL FOR WATER PUMPS STATUS TO IMPROVE EFFICIENCY ON DELIVERY IN TANZANIA KEY RESULTS Identified the functionality of the pumps The model achieved 81% accuracy The data will enable better decision making and service delivery PERFORMANCE OF THE MODEL Model Used: Random Forest Classifier Accuracy: 81% Key Metrics Observed Fully Fuctional Pumps – F1: 0.85 | Precision: 0.81 | Recall: 0.89 Functional But Need repairs - F1: 0.41 | Precision: 0.55 | Recall: 0.33 Non-Functional - F1: 0.81 | Precision: 0.84 | Recall: 0.78 Conclusion: The class of pumps that are functioning but need repairs is the hardest to predict Key takeaways for the Management Company A significant number of non-functional pumps might have been prevented through timely intervention & preventive maintenance. Key Predictors of Pump Failure: Construction Year – Older pumps show higher likelihood of failure Region – Geographic disparities influence functionality Management Type – Community or institutional oversight impacts performance Installer – Quality and consistency vary by installation provider The model enables proactive identification of high-risk pumps, allowing stakeholders to forecast failures before they occur and prioritize preventive maintenance Recommendations Proactive Maintenance - Prioritize inspections and repairs for pumps flagged as high-risk by the model. Targeted Investment - Direct funding and resources to regions with high failure rates to maximize impact. Enhanced Data Collection - Train field teams and standardize reporting formats to improve data quality and model accuracy. Digital Monitoring - Integrate sensors or mobile applications for real-time tracking and remote diagnostics. Stakeholder Dashboard - Develop an interactive platform to visualize pump functionality, risk levels, and predictive insights. Impact Potential Short-Term Benefits: Reduced Downtime: Minimize pump outages and mitigate water scarcity Optimized Maintenance: Enable efficien …