IPMI experts predict changes in core and seasonal inflation variables for 2023

Tashkent, Uzbekistan (UzDaily.com) -- Experts from the Institute for Forecasting and Macroeconomic Research (IPMI) predicted changes in the base and seasonal inflation variable (prices for fruits and vegetables) for 2023.

For this, the ARIMA and SARIMA models were applied. The complexity of the task of forecasting time series lies in the temporal variability of the series and the inability of forecasting models to fully reflect these changes. To ensure that changes over time are automatically accounted for, IPMI has developed a framework to optimize the relative ordering of models used in time series forecasting.

The components of inflation were divided into 3 main groups, which were then analyzed and forecasted. These include: core inflation, managed prices, and inflation with seasonally variable prices (fruits and vegetables). The calculation of core inflation does not take into account the types of goods and services that are affected by seasonal changes. Regulated price inflation consists of administratively regulated goods and services. When calculating the seasonal inflation variable, prices for vegetables and fruits are mainly taken into account.

In the case of Uzbekistan, 78% growth in the consumer price index (CPI) corresponds to core inflation. Due to this, monetary policy decisions are usually made based on the results of this inflation.

Fruits and vegetables are the most volatile groups in the consumer price index (CPI) basket, accounting for 8% of the total FPI basket.

According to the results of the IPMI forecast, core inflation for 2022 is 11.2%, and horticultural inflation is 15.9%. For 2023, core inflation is expected at 10.5%, and horticultural inflation at 22.1%.

The proposed method dynamically optimizes the results of the model (meta-learning). The optimization is repeated until the most suitable model based on the time series function is found (iterative method).

In conclusion, testing of individual time series forecasting models is time-consuming and should be done taking into account the specifics of the predicted data. Therefore, IPMI experts will continue forecasting time series using machine learning algorithms (machine learning framework).