Factor-Based Framework for Multivariate and Multi-step-ahead Forecasting of Large Scale Time Series

State-of-the-art multivariate forecasting methods are restricted to low dimensional tasks, linear dependencies and short horizons. The technological advances (notably the Big data revolution) are instead shifting the focus to problems characterized by a large number of variables, non-linear dependencies and long forecasting horizons. In the last few years, the majority of the best performing techniques for multivariate forecasting have been based on deep-learning models. However, such models are characterized by high requirements in terms of data availability and computational resources and suffer from a lack of interpretability. To cope with the limitations of these methods, we propose an extension to the DFML framework, a hybrid forecasting technique inspired by the Dynamic Factor Model (DFM) approach, a successful forecasting methodology in econometrics. This extension improves the capabilities of the DFM approach, by implementing and assessing both linear and non-linear factor estimation techniques as well as model-driven and data-driven factor forecasting techniques. We assess several method integrations within the DFML, and we show that the proposed technique provides competitive results both in terms of forecasting accuracy and computational efficiency on multiple very large-scale (>102 variables and > 103 samples) real forecasting tasks.

Model and methods of multi-variant forecasting the state of complex technical objects in proactive control of them

The article shows that in proactive management of complex technical objects, it is very important to carry out multivariate forecasting of their state. A generalized model and methods for carrying out the specifi ed forecasting are proposed, based, fi rstly, on the original logical-dynamic description of the processes under study, and, secondly, on the construction and approximation of the reachable regions of the investigated dynamic processes that defi ne the entire set of potential trajectories of SRT movement in the multidimensional space of system-technical parameters.

Implementasi Model Fungsi Transfer dan Neural Network untuk Meramalkan Harga Penutupan Saham (Close Price)

The multivariate forecasting model is a model of forecasting that takes into the causal relationship between a prediction factor with one or more independent variables. This study uses multivariate forecasting model that are transfer function and neural network model. The transfer function and neural network model are used for forecasting of closing stock price data by considering the opening stock price data as the independent variable in the forecasting model. The data used in this study is the monthly closing stock price and opening stock price data of PT. Bank Central Asia, Tbk. The best model for forecasting of closing stock price is a transfer function model that has MSE, MAPE, and MAE values ??smaller than the neural network model. Keywords: transfer function, neural network, opening stock price, closing stock price