An Intelligent Hybrid Model Using Artificial Neural Networks and Particle Swarm Optimization Technique For Financial Crisis Prediction

Abstract

Financial crisis prediction is a critical issue in the economic phenomenon. Correct predictions can provide the knowledge for stakeholders to make policies to preserve and increase economic stability. Several approaches for predicting the financial crisis have been developed. However, the classification model's performance and prediction accuracy, as well as legal data, are insufficient for usage in real applications. So that, an efficient prediction model is required for higher performance results. This paper adopts a novel two-hybrid intelligent prediction model using an Artificial Neural Network (ANN) for prediction and Particle Swarm Optimization (PSO) for optimization. At first, a PSO technique produces the hyperparameter value for ANN to fit the best architecture. They are weights and thresholds. Then, they are used to predict the performance of the given dataset. In the end, ANN-PSO generates predictions value of crisis conditions. The proposed ANN-PSO model is implemented on time series data of economic conditions in Indonesia. Dataset was obtained from International Monetary Fund and the Indonesian Economic and Financial Statistics. Independent variable data using 13 potential indicators, namely imports, exports, trade exchange rates, foreign exchange reserves, the composite stock price index, real exchange rates, real deposit rates, bank deposits, loan and deposit interest rates, the difference between the real BI rate and the real FED rate, the M1, M2 multiplier, and the ratio of M2 to foreign exchange reserves. Meanwhile, the dependent variable uses the perfect signal value based on the Financial Pressure Index. A detailed statistical analysis of the dataset is also given by threshold value to convey crisis conditions. Experimental analysis shows that the proposed model is reliable based on the different evaluation criteria. The case studies show that the result for predictive data is basically consistent with the actual situation, which has greatly helped the prediction of a financial crisis.

RETRACTED ARTICLE: Explainable AI Model for Recognizing Financial Crisis Roots Based on Pigeon Optimization and Gradient Boosting Model

“…Swarm optimization approaches such as Ant Colony Optimization (ACO) [ 31 ], Particle Swarm Optimization [ 32 ], Grey Wolf Optimization (IGWO), and Fuzzy Neural Classifier (FNC) [ 33 ] can be used also to develop smart systems able to accurately predict the financial crisis and bankruptcy. Although many appreciated efforts have been paid to solve financial crisis prediction issues, few works have used the concept of Explainable Artificial Intelligence (XAI) in predicting financial crises.…”

Classification of Solo Batik patterns using deep learning convolutional neural networks algorithm

“…CNN is based on the artificial neural networks (ANN) architecture. The main objective of designing and training a neural network model is to identify, modify, and optimize parameters in order to minimize objective functions and generate precise predictions for incoming inputs [16].…”

Predicting Models of Financial Crises

“…Different features of the financial crisis have been studied by economists in order to identify an accurate methodology for predicting financial crisis using the data from historical years [1,2]. The prediction aims to forewarn the decision-makers within the economy to prevent or reduce the damage caused by the crisis as much as possible by taking advanced actions.…”