Big Data Anlaysis on Slope Repair Request

The project aimed to leverage machine learning, specifically the Random Forest model, for big data analysis to predict the occurrence of Slope Repair Requests (SRR) on a list of over 2700 slopes. The input dataset encompassed diverse factors such as slope geometry, location, past SRR history, vegetation details, maintenance parties involved, meteorological data, and more.

Random Forest is an ensemble learning technique that builds multiple decision trees and merges their predictions. Each decision tree is constructed based on a random subset of the input features, and the final prediction is determined by the consensus of all trees. This approach not only minimizes overfitting but also provides robust predictions for diverse datasets.

The process involved comprehensive data preprocessing, including cleaning, normalization, and feature engineering. Robotic Process Automation was depolyed to collect data from government website. The Random Forest algorithm was chosen due to its ability to handle large datasets with numerous features. Optimization was conducted by tuning hyperparameters such as the number of decision trees, maximum tree branches, and the amount of data in each leaf of the tree. This meticulous optimization aimed to enhance the model's accuracy and efficiency.

The model was tested using unseen data to evaluate its predictive capabilities. Remarkably, it demonstrated the ability to predict over 70% of SRR cases with only 30% of the effort traditionally required. Subsequently, a ranking of each slope was produced, facilitating the prioritization of preventive maintenance efforts.Based on the model's predictions, extra preventive maintenance works were conducted proactively on high-risk slopes. As a result, the occurrence of SRR reduced by over 30%. This not only averted potential hazards but also resulted in substantial cost savings, exceeding 2 million HKD. The return on investment (ROI) for the project exceeded 400%, showcasing the efficacy of leveraging machine learning for infrastructure maintenance.