Scholarlink

This research paper examines the efficiency of Deep Learning models of LSTM and N-BEATS, and the statistical methods of ARIMA and VAR alongside the impact of macroeconomic factors-specifically the Consumer Price Index (CPI) and Interest Rates-on Microsoft’s stock prices from August 2009 to June 2020. The paper also reviews a variety of sources that analyze the statistical relationships between macroeconomic factors and the stock market, highlighting these models’ advantages for time series forecasting. While ARIMA, N-BEATS, LSTM, and VAR are recognized as top models for time series predictions, their performance varies with data characteristics, model architecture, and tuning. This paper provides a detailed explanation of the concepts and mathematical formulations behind each implemented algorithm, making it a valuable resource. In examining univariate forecasting (ARIMA, LSTM, N-BEATS) and multivariate forecasting (VAR), it’s seen that incorporating macroeconomic indicators in multivariate forecasting resulted in lower MSE and MAPE. Specifically, VAR achieved the lowest MSE of 59.5901 and MAPE of 0.0476, proving it the most efficient model among those tested.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Analysing the Variation Between Univariate Forecasting and Multivariate Forecasting Using Macroeconomic Factors, for Stock Price Prediction

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Abha Barge,
  • Sandhya Arora

摘要

This research paper examines the efficiency of Deep Learning models of LSTM and N-BEATS, and the statistical methods of ARIMA and VAR alongside the impact of macroeconomic factors-specifically the Consumer Price Index (CPI) and Interest Rates-on Microsoft’s stock prices from August 2009 to June 2020. The paper also reviews a variety of sources that analyze the statistical relationships between macroeconomic factors and the stock market, highlighting these models’ advantages for time series forecasting. While ARIMA, N-BEATS, LSTM, and VAR are recognized as top models for time series predictions, their performance varies with data characteristics, model architecture, and tuning. This paper provides a detailed explanation of the concepts and mathematical formulations behind each implemented algorithm, making it a valuable resource. In examining univariate forecasting (ARIMA, LSTM, N-BEATS) and multivariate forecasting (VAR), it’s seen that incorporating macroeconomic indicators in multivariate forecasting resulted in lower MSE and MAPE. Specifically, VAR achieved the lowest MSE of 59.5901 and MAPE of 0.0476, proving it the most efficient model among those tested.