Systems for decision-making assistance are gradually using analytical and computational methods to aid in management as well as decisions regarding strategy. In order to use these kinds of technologies to reliably predict economic information, researchers need to understand how to use them. Consequently, this chapter presents a method-based literature assessment with a focus on the subject of predictive analytics. The study covers the time series simulations, association, regression analysis, grouping, and categorization in great detail. It introduces machine learning into the realm of mathematical explanatory modeling. The approaches examined enable future prediction through the analysis of longitudinal and financial time series data collected, preserved, and handled in computer systems. The outcomes of these models aid in improving outcomes for risk administration specialists and financial executives. This review unifies several financial forecast analytic methodologies into a single domain.
Top1. Introduction
The application of predictive analytic techniques by academics in the disciplines of finance and computation is reviewed and examined in this work. Data from financial Information Systems (IS), whether internal or external, is especially used to identify systematic approaches and use cases. The modelling process itself is not explained in this work; it only serves as a review. Many textbooks provide adequate coverage of this. Rather, it is an aggregate collection of mathematical and analytical justifications for the business of finance that were obtained through bibliometric and keyword searches. Given that forecasting approaches are being used by DSS (Decision Support Systems) more and more, it is relevant to the IS area. In the corporate realm, the outcomes are used to boost revenues or the efficiency of products.
The word “finance” encompasses a wide range of financial market operations, such as intertemporal and managing portfolios. (Summers, L. H. 1985). Time series are typically shaped by these activities. The future values and/or returns of these can be predicted using these time series (De Gooijer, J. G., & Hyndman, R. J. 2006). Weak normality assumptions in the means, variances, and covariances of these time series serve as the basis for statistical inference. Additionally, it can be found in cross-sectional series produced by markets and financial activities. They serve as the foundation for analysis based on stats and can be incorporated into a DSS to forecast consumer, business, or information for the industry as well. Cross-sectional analysis's use in predicting a company's beta is one example. This shows the systematic risk of the company concerning the market and is generated from the capital asset pricing model (CAPM). The slope coefficient, or beta, is easily determined by using a least squares regression.
Transforming from descriptive arithmetical models to computing forecasts, the concept of predictive analytics is more frequently used. The latter is a requirement for DSS, while the former is increasingly incorporated into them. The latter's algorithmic methods evaluate the validity and accuracy of its results using test data sets. Contrarily, confidence intervals and significance tests are used in statistical models. That being said, both computational and statistical methods are discussed in this review because they are combined in data analytics.
In the banking industry, statistical analysis is used to assess past IS data sets. The term encompasses both descriptive analytics and prognostic mathematical techniques. The findings of an approach using an algorithm are obtained from the data instead of human interpretation, in contrast to the usual application of forecasting in finance. The authors caution those who are using this technique as a computation for prediction to be at ease with the notion that data, not theory, should be the starting point of their work. (Frecka, T. J., & Hopwood, W. S. 1983); however, caution that these same researchers should not exercise caution to avoid over-fitting, where the enticement may exist to develop a concept based on the datasets and formerly manipulate the datasets to support the hypothesis.
Data ought to be DSS needs comprehensive and clean data from an organization's computer system in order to execute analytics that are predictive. The vast, varied, and multidimensional nature of statistics makes it difficult to handle the numerous, autonomous, multifarious, and ever-changing “Big Data” that is commonly observed in the financial industry. (Wu, X. et al, 2013)
It raises a vital question for IS. It's critical to make the distinction between data and information created internally and externally. Therefore, Section 2 of this article covers the datasets that need to be loaded into an IS, while Section 3 covers the datasets generated by an IS. To handle methods of internally gathered data in its raw form, a collaborative software-driven solution is required. To handle methods to external data, a system for gathering dispersed data or relevant imported sets of information is required.