A Novel Ensemble Model by Combining LSTM, BiLSTM, and Facebook Prophet Algorithm to Forecast Stock Prices

Abstract

The nature of stock prices is very volatile. The data is fluctuated and is updated every second due to various factors like news, economic, and social responses. This volatility results in extensive data, which is challenging for traditional financial strategies to predict stocks' closing values. Machine Learning (ML) and Deep Learning (DL) predictive models can process extensive data, visualize, and forecast the results accurately within a short time. Past researchers have proposed various ML and DL predictive models. We observed that the usage of Long Short-term Memory (LSTM) and the Facebook Prophet algorithm is trending in forecasting time-series data. After exploring the types of LSTM models, Bidirectional LSTM (BiLSTM) proves to be a robust predictive model. BiLSTM is an improved version of LSTM. FB Prophet algorithm provides realistic predictions by handling the seasonality and detecting trend variation in extensive data. Our body of work in this research highlights the contemporary predictive models and proposes a novel ensemble model of LSTM, BiLSTM, and FB Prophet algorithm. The novelty of this work is to combine three robust predictive models using the weighted average technique and compensate for the limitations of one model with the advantages of the other to provide more reliable forecasts of Google stock prices.