191 LSTM-BASED AUTOMATED LEARNING WITH SMART DATA TO IMPROVE MARKETING FRAUD DETECTION AND FINANCIAL FORECASTING Zair Bouzidi 1 Abdelmalek Boudries 2 Mourad Amad 3 DOI: https://doi.org/10.31410/EMAN.2021.191 Abstract: This proposed model is based on a deep recurrent neural network trained with Long Short- Term Memory Network (LSTM), used because of its ability to learn long term dependencies, taking the concatenated function and Financial data as input, while integrating encapsulations, using Smart Data and retrieving information by combining multiple search results (all the Web). It combines representa- tion training with fnancial data while integrating encapsulations from multiple sources and retrieving information by combining multiple search results. It provides some good ideas that we have extended to improve Corporate Marketing and Business Strategies. We show that the proposed model learns to localize and recognize diferent aspects of Corporate Marketing and Business Strategies. We evaluate it on the challenging task of detecting Fraud in Financial Services and Financial Time Series Forecast- ing and show that it is more accurate than the state-of-the-art of other neural networks and that it uses fewer parameters and less computation. Keywords: Business, Marketing, Forecasting of fnancial times series, Fraud detection, LSTM, Smart Data. 1. INTRODUCTION T he prime goal of a fnancial time series model is to provide reliable future forecasts which are crucial for investment planning, fscal risk hedging, governmental policy making, etc. These time series often exhibit notoriously haphazard movements which make the task of mod- eling and forecasting extremely difcult. As per the research evidence, the random walk (RW) (Fama, 1995) is so far the best linear model for forecasting fnancial data. Artifcial neural net- work (ANN) is another promising alternative with the unique capability of nonlinear self-adap- tive modeling. Numerous comparisons of the performances of RW and ANN models have also been carried out in the literature with mixed conclusions (Adhikari, 2014). We propose a new real-time automated learning model based on a recurrent neural network trained with Long Short-Term Memory Network (LSTM) that integrates encapsulations using Smart Data and thus retrieves information by combining multiple search results from multiple sources (all the Web). Thus, we provide not only a solution to this challenge, but also, propose better performances. 1 LIMPAF Laboratory, Computer Science Dept, Faculty of Science and Applied Science, Bouira University, Algeria 2 Laboratory LMA, Commercial Science Dept, Faculty of Economics, Business and Management, Bejaia University, Algeria 3 LIMPAF Laboratory, Computer Science Dept, Faculty of Science and Applied Science, Bouira University, Algeria