Prediction of Fund Net Value Based on ARIMA-LSTM Hybrid Model

Abstract

The net value of fund is affected in many ways, and researchers attempt to quantify these influences in order to predict future net value by developing various models. Current prediction models typically can only reflect the linear variation law, and their nonlinear characteristics are either poorly handled or selectively ignored, resulting in less accurate prediction results. Based on this, the ARIMA-LSTM hybrid model is used in this paper to predict funds. After preprocessing historical data, the ARIMA model is used to filter out the linear data characteristics, followed by the LSTM model to extract the nonlinear characteristics by residual, and finally superposition the respective prediction values of the two models was performed to obtain the hybrid model's prediction results. The paper's methodologies are empirically proven to be more accurate and applicable than typical fund forecast methods.

Influence of Artificial Intelligence on Forecasting Net Asset Value and Return Volatility in Indian Mutual Fund

“…Conversely, the encoder-decoder convolutional LSTM model finds that its forecasting results are most accurate when it takes the data from the prior two weeks as the input is the fastest in its execution. In [7], the authors created an ARIMA-LSTM hybrid model based on a 1260-daily NAV from June 2016 until July 2021. The hybrid ARIMA-LSTM model is the basis of this paper's fund prediction technique.…”