Farid Abdalla

Summary

The PPI Doppelganger project aimed to develop a deep learning model that predicts the trends of currency pairs based on historical data. The model's primary objective was to provide reliable and transparent predictions that assist professional traders in making informed decisions.

Our approach utilized various features such as previous OHLC data, inflation index, and gold prices. The model was built using TensorFlow and Keras libraries, and explainability was a critical focus. We employed SHAP and LIME techniques to ensure that the model's decision-making process was transparent and understandable for professional traders.

Despite the best efforts to optimize the model's performance, the best accuracy we achieved was around 55%. Furthermore, the model required frequent retraining to maintain consistency, which was a troublesome and time-consuming process. To address these challenges, the client decided to shift the project's focus towards NLP analysis.

The revised project now includes text classification, document summarization, topic modeling, and sentiment analysis. The objective is to help traders make informed decisions by analyzing various documents, such as bank reports and news articles, and extracting relevant information.