Improving Time Series Prediction With Feature Selection Using a Velocity-Enhanced Whale Optimization Algorithm

1. Introduction

Time series data are noisy and non-stationary. Due to the minimum availability of data, these time series data act as noisy and time series distribution changes with time that makes it non-stationary (Lu et al., 2009). Both the economy of a developing country and societal comfort rely on performance of financial section of the country (Deboeck, 1994). The growing need of individuals to access basic amenities like education and health as well as the need to reduce poverty has motivated to study the behaviour of the financial time series data with the help of different type of soft computing techniques (Yaser & Atiya, 1996)

To cope with the future dynamic environment of finance, the development of an evolutionary approach is both a necessary and an important task (Bahrammirzaee, 2010). Thus, this paper provides a significant model to describe financial prediction using a nature-based meta-heuristic approach. The objective of this paper includes: (i) adapt to a nature-based meta-heuristic approach to develop a financial prediction model, (ii) financial and non-financial ratios are used to rise the accuracy of the model, (iii) some traditional techniques are applied to compare the degree of accuracy with that of AI approach, (iv) this model is expanded such that it will work within financial prediction system in order to give information to investment monitoring organization as well as investors.

Related works are discussed in Section 2. VEWOA-NN is explained in Section 3. Section 4 explains about different types of prediction algorithms. A Summary of feature selection methods is debated in Section 5. Section 6 results are presented, and Section 7 defines the Conclusion.