Enhancing Fare Prediction Accuracy in Ride Hailing Through Neural Networks and Data Simulation

Prediction of fare is the need for better service quality and effective pricing strategies that make for enhanced customer experiences in a ride-hailing service. From the identified features based on rides and ride-hailing activities - the distance covered, the time used for travel, number of passengers, and the main purposes of trips, the researchers will create a neural network model using such synthesized data to make real-time predictions for an accurate total fare estimate. This synthesis helps to understand the noise and natural variability in the ride-hailing datasets by showing different interaction patterns of the features over time. The analysis techniques used include scatter plots, correlation matrices, and time-series evaluations that reveal strong connections between fare amounts and distance traveled and trip duration. The dataset used in this research was sourced from Kaggle, ensuring a reliable foundation for training and evaluating the model. The proposed neural network architecture achieved 90% accuracy, with Mean Absolute Error (MAE) of 1.5 units and Root Mean Squared Error (RMSE) of 2.3 units, demonstrating its reliability. This paper shows how neural networks might be applicable to fare predictions and, therefore, can contribute toward artificial datasets with practical models applicable in the real world.