Large Dataset: The project dealt with a substantial dataset of 3.6 million rows of unstructured text data from Amazon reviews, requiring robust natural language processing (NLP) techniques to analyze sentiments.
Sentiment Analysis: The primary objective was to determine whether reviews for a product were positive or negative. NLP methods were employed to understand the sentiment behind each review, helping in categorizing them accurately.
LSTM Algorithm: Long Short-Term Memory (LSTM) was the chosen algorithm for this task. LSTMs are recurrent neural networks well-suited for sequential data like text, making them effective for sentiment analysis.
Machine Learning Training: The project involved training the LSTM model on the historical dataset to learn the patterns and sentiments expressed in the reviews. This trained model was then used for future predictions.
Predictive Analytics: The project's outcome enables predictive analytics, allowing for real-time or batch processing of new reviews to automatically categorize them as positive or negative, providing valuable insights for product evaluation and improvement.