Forecasting and Technical Comparison of Inflation in Turkey With Box-Jenkins (ARIMA) Models and the Artificial Neural Network

Inflation refers to an ongoing and overall comprehensive increase in the overall level of goods and services price in the economy. Today, inflation, which is attempted to be kept under control by central banks or, in the same way, whose price stability is attempted, consists of continuous price changes that occur in all the goods and services used by the consumers. Undoubtedly, in terms of economy, in addition to the realized inflation, inflation expectations are also gaining importance. This situation requires forecasting the future rates of inflation. Therefore, reliable forecasting of the future rates of inflation in a country will determine the policies to be applied by the decision-makers in the economy. The aim of this study is to predict inflation in the next period based on the consumer price index (CPI) data with two alternative techniques and to examine the predictive performance of these two techniques comparatively. Thus, the first of the two main objectives of the study are to forecast the future rates of inflation with two alternative techniques, while the second is to compare the two techniques with respect to statistical and econometric criteria and determine which technique performs better in comparison. In this context, the 9-month inflation in April-December 2019 was forecast by Box-Jenkins (ARIMA) models and Artificial Neural Networks (ANN), using the CPI data which consist of 207 data from January 2002 to March 2019 and the predictive performance of both techniques was examined comparatively. It was observed that the results obtained from both techniques were close to each other.

On the applicability of ARPS models in modelingand forecasting the dynamics of new cases of COVID-19 infection

The aim of the work is to study the applicability of the methodology for constructing ARIMA models in the problem of modeling and predicting the dynamics of new cases of coronavirus infection in the autumn-winter period of 2020 in the KhMAO-Yugra.

ARIMA model for forecasting of evaporation of Solapur station of Maharashtra, India

This paper deals with the stochastic modeling of weekly evaporation by using Seasonal ARIMA model for weekly evaporation data for the period of 1987-2008 with a total of 1144 readings for semi-arid Solapur station in Maharashtra. ARIMA models of 1st order were selected based on observing autocorrelation function (ACF) and partial autocorrelation function (PACF) of the weekly evaporation series. The model parameters were obtained by using maximum likelihood method with the help of three tests (i.e., standard error, ACF and PACF of residuals and Akaike Information Criteria). Adequacy of the selected models was determined. The ARIMA model that passed the adequacy test was selected for forecasting. The Seasonal ARIMA (1, 0, 1) (1, 0, 1)52 with lower RMSE is finally selected for forecasting of weekly evaporation values, at Solapur.

ARIMA Models in Electrical Load Forecasting and Their Robustness to Noise

The paper addresses the problem of insufficient knowledge on the impact of noise on the auto-regressive integrated moving average (ARIMA) model identification. The work offers a simulation-based solution to the analysis of the tolerance to noise of ARIMA models in electrical load forecasting. In the study, an idealized ARIMA model obtained from real load data of the Polish power system was disturbed by noise of different levels. The model was then re-identified, its parameters were estimated, and new forecasts were calculated. The experiment allowed us to evaluate the robustness of ARIMA models to noise in their ability to predict electrical load time series. It could be concluded that the reaction of the ARIMA model to random disturbances of the modeled time series was relatively weak. The limiting noise level at which the forecasting ability of the model collapsed was determined. The results highlight the key role of the data preprocessing stage in data mining and learning. They contribute to more accurate decision making in an uncertain environment, help to shape energy policy, and have implications for the sustainability and reliability of power systems.

Price-limit effectiveness: evidence from the Borsa Istanbul (BIST)

Purpose This paper aims to analyze the impact of price-limit hits by hit type and when such hits start and stop using intraday trades and quotes at a one-second frequency for firms included in the BIST-50 index during the 13-months starting with March 2008. Like the recent COVID-19 period, this period includes the heightened stress in global financial markets in September 2008. Design/methodology/approach Using intra-day trades and quotes at a one-second frequency, the authors examine the market effects of price limits for firms included in the BIST-50 index during the global financial crisis. The authors compare the values of various metrics for 60 min centered on price-limit hit periods. The authors conduct robustness tests using auto regressive integrated moving average (ARIMA) models with trade-by-trade and with 3-min returns. Findings The findings are supportive of the following hypotheses: magnet price effects, greater informational asymmetric effects of market quality and each version of price discovery. Results are robust using samples differentiated by cross-listed status, same-day quotes instead of transaction prices and equidistant and trade-by-trade returns. Originality/value The authors use intraday data to reduce measurement error that is particularly pronounced when daily data are used to assess price limits that start and/or stop during a trading session. The authors contribute to the micro-structure literature by using ARIMA models with trade-by-trade and 3-min returns to alleviate some bias due to the autocorrelations in returns around price-limit hits in the presence of a magnet effect. The authors include some recent regulation changes in various countries to illustrate the importance of circuit breakers using price limits during COVID-19.

Determining the Rolling Window Size of Deep Neural Network Based Models on Time Series Forecasting

Time series forecasting has always been a significant task in various domains. In this paper, we propose DeepARMA, a LSTM-based recurrent neural network to tackle this problem. DeepARMA is derived from an existing time series forecasting baseline, DeepAR, overcoming two of its weaknesses: (1) rolling window size determination: the way DeepAR determines rolling window size is casual and vulnerable, which may lead to the unnecessary computation and inefficiency of the model;(2) neglect of the noise: pure autoregressive model cannot deal with the condition where data are composed of various kinds of noise, neither do most of time series models including DeepAR. In order to solve these two problems, we first combine a classic information theoretic criterion, AIC, with the network to determine the proper rolling window size. Then, we propose a jointly-learned neural network fusing white Gaussian noise series given by ARIMA models to DeepAR’s input. That is exactly why we name the network ‘DeepARMA’. Our experiments on a real-world dataset demonstrate that our improvement settles those two problems put forward above.