TRADITIONAL ARIMA, LSTM AND HYBRID TECHNIQUES FOR ACCURATE PLATINUM PRICE PRIDICTION

Abstract

Time series analysis becomes a vital tool in engineering, finance, and social research. Originally, univariate Autoregressive (AR) and univariate Moving Average (MA), Simple Exponential Smoothing (SES) model was developed to forecast next period data. Additionally, ARIMA was developed to deal with nonstationarity data. Specifically, the ARIMA model has shown superior accuracy and precision in forecasting the upcoming time series lags. Later, Artificial Neural Network (ANN) model and Long Short-Term Memory (LSTM) model are widely used in time series analysis for the different research areas. For predicting platinum prices, conventional mathematical model ARIMA and a non-linear method LSTM have been developed . Hybrid model has introduced in addition to LSTM model and ARIMA, conventional time series models.The primary goal of this study is to examine Hybrid capacity for modeling variations in the price of platinum and to assess how well it performs in comparison to other established time series modeling methods like ARIMA. Finally, based on performance standards including Mean Absolute Error(MAE), Root Mean Square Error(RMSE) the best-fit model is determined. Further the percentage better performance of the model is applied to test the accuracy of these models. The
UtilitasMathematica
ISSN 0315-3681 Volume 122, 2025
177
findings demonstrate that Hybrid technique is a potent tool for modeling the platinum price and can provide more accurate forecasts than LSTM and ARIMA model.