The Decomposition and Forecasting of Mutual Investment Funds Using Singular Spectrum Analysis

Area of Science:

• Time Series Analysis
• Statistical Modeling
• Data Science

Background:

• Singular Spectrum Analysis (SSA) is a powerful non-parametric technique for decomposing time series into interpretable components like trend, periodicity, and noise.
• Traditional SSA and other statistical methods can yield inaccurate results when time series data contains outliers.
• Robust methodologies are essential for reliable analysis in the presence of data anomalies.

Purpose of the Study:

• To evaluate the performance of robust SSA algorithms against classical SSA and ARIMA models.
• To compare the accuracy and computational efficiency of these methods for time series model fitting and forecasting.
• To demonstrate the superiority of robust SSA in handling time series data with outliers.

Main Methods:

• Implementation of two robust SSA algorithms for model fitting and one for forecasting.
• Comparison with the classic SSA model and the Autoregressive Integrated Moving Average (ARIMA) model.
• Validation using both simulated data and real-world time series data from mutual investment funds.

Main Results:

• Robust SSA algorithms significantly outperform classical ARIMA and SSA models when outliers are present in the data.
• The simulation study confirmed the enhanced accuracy of robust SSA methods in the presence of anomalies.
• Performance was assessed based on model fit accuracy, forecast accuracy, and computational time.

Conclusions:

• Robust SSA algorithms provide a more reliable approach for time series analysis and forecasting, especially when dealing with outlier-corrupted data.
• The findings highlight the importance of employing robust statistical methods for accurate data interpretation and prediction.
• This study advocates for the adoption of robust SSA techniques in fields susceptible to data anomalies.