Evaluation of Financial Data Processing Life Cycle for Risk Prediction: A Survey

Financial risk analysis is integral to financial planning and investment at organizational and personal levels. Due to the higher fluctuation of the financial trend, many inverters consider the risk prediction strategy during the investment portfolio generation. The risk prediction for financial assets is highly challenging due to the dependency of financial trends on various technical and non-technical factors. Hence, the use of computer-aided processes is becoming popular for risk prediction. Recently, with the enhancements in machine learning algorithms, the risk prediction processes have improved the accuracy of the prediction. These algorithms have two phases: training the model and deploying the models to predict. Nonetheless, the available machine learning algorithms for risk prediction have many limitations. The limitations primarily concern the correctness of the data to be deployed for building the predictive model for prediction as these data are collected from various sources, sometimes with human interventions, and are prone to insufficient and incorrectness. Hence, the frameworks or the processes for financial predictions must perform an additional step, such as data pre-processing, and then further perform the actual task, risk predictions. In the recent past, a good number of research works have aimed to predict financial risks with higher accuracy by designing a complete life cycle of the data for financial predictions, starting from data pre-processing to the conclusion of risk analysis. Nevertheless, these works are criticized for not performing the prediction task with the best possible accuracy and compromising on the time complexity, as time complexity can be a critical measure of performance in financial risk analysis. Hence, this work aims to analyze the various strategies and works for data pre-processing and predictions on financial data. This work finally contributes to the research domain by analyzing the strategies mathematically, algorithmically and result wise to identify the unsolved challenges in this domain.