ARCHIVES

Ride-hailing platforms such as Ola, Uber, Rapido, Lyft, Bolt, Grab, and other app-based transport services have become an important part of modern urban mobility. Customer reviews posted on these platforms contain useful opinions about service quality, driver behavior, booking experience, pricing, cancellation issues, payment problems, and overall user satisfaction. Since manual ana lysis of large review data is difficult, automated sentiment analysis is useful for understanding customer feedback. This project presents a deep learning-based binary sentiment analysis framework for classifying ride-hailing customer reviews into positive and negative categories. The original dataset contained 12,000 reviews with review text and rating scores. After removing neutral ratings, the final dataset contained 11,667 reviews, including 8,732 positive reviews and 2,935 negative reviews. Reviews with ratings 4 and 5 were labelled as positive, while ratings 1 and 2 were labelled as negative. The review text was cleaned by applying lowercase conversion, URL removal, special character removal, stopword removal, tokenization, and sequence padding. The proposed model was developed using a Bidirectional Long Short-Term Memory network. The model includes an embedding layer, SpatialDropout1D layer, BiLSTM layer, dropout layer, dense layer, and sigmoid output layer for binary classification. The dataset was divided into 80% training data and 20% testing data. The model was trained using binary cross-entropy loss and Adam optimizer with early stopping and learning-rate reduction.