Forecast methods offers 10 forecasting fully automated methods.
Moving averages are the methods which flexibly adjust to the analysed data thus it is easier for them to foresee any changes in the overall economic situation on the market. Moving averages work well in forecasting prices of agricultural products as well as in forecasting stock exchange quotations.
The following moving averages are accessible at platform:
  • SMA (Simple Moving Average)
  • WMA (Weighted Moving Average)
  • EMA (Exponential Moving Average)
  • AMA (Adaptive Moving Average)
The abovementioned moving averages available at the platform require the following minimum number of observations:
  • SMA, WMA and EMA: 4 observations,
  • AMA: 6 observations.
Moving averages are available for Basic Account users.
The Holt Model, similarly to the moving averages, is able to flexibly address trends in changes and constantly correcting its matching. The Holt Model stacks up well for forecasting non-seasonal time data therefore it is not available on the platform for all datasets.
Please note that generating a forecast based on the Holt model requires at least 4 observations.
The Winters Model constitutes an extension to the Holt model by taking seasonality into consideration. It works especially well for quarterly and monthly data, such as monthly sale of products or quarterly number of orders. The Winters method is available only for monthly and quarterly data.
In the system, the Winters model forecasting will require at least:
  • 8 observations for quarterly data,
  • 24 observations for monthly data.
Trend is a curve of a gradual and flat course describing long-term changes and presenting the main direction of development. Thanks to the trend forecasting we can foresee how the values of our data might develop in a longer period of time (e.g. sales) in upcoming months. There are seven types of trends available on
  • linear
  • quadratic
  • cubic
  • hyperbolic
  • logarithmic
  • root
  • exponential
In our system, the model of trend is chosen automatically based on the following criteria: how it matches the data and also based on reliability of the forecast. The user selects a trend as the forecast method only, whereas its type gets selected automatically based on the reliability of results.
5 observations are required for the forecasting based on trends.
The Component Model is a tool which allows to grasp both a trend and seasonality simultaneously. It includes trend, seasonality and a component providing historical data. It is less flexible than moving averages model but it provides more reliable forecasts, especially for the long-term ones. It is a much more advanced tool for the forecasting than methods discussed above because it uses more advanced theories and calculation techniques. The Component model is used for illustrating the phenomena which demonstrate either increasing or decreasing tendencies and can change seasonally (daily, monthly or quarterly), e.g. building materials output or agricultural goods output.
In the system, there are two types of the componential model used for the forecasting purposes and those are automatically chosen by the program (depending on the type of the forecasted data):
  • T-AR – Trend-Autoregressive requiring at least 6 observation,
  • T-S-AR – the Trend-Seasonal-Autoregressive model; its forecasting results require at least 12 observations for the quarterly data and at least 28 observations for the monthly data.
At the platform the optimal form of the component model gets selected automatically from the maximum 392 variants.
The SARIMA model is the most advanced forecasting model among all the available methods in our system. This method is used, among others, by the European Statistical Office “Eurostat” and also by US Census Bureau. It considers various types of trends as well as different variations of seasonality. Additionally, it is largely based on historical data therefore it is very universal. At an optimal type of the SARIMA model is selected automatically from 128 possible combinations.
In order to generate forecasts using the SARIMA method the following minimum number of observations are required:
  • 12 for daily, weekly and annual data,
  • 14 for quarterly data,
  • 24 for monthly data.
The Expert method is the most technically advanced forecast method available in our system. It is a method which optimally combines the effects of all the methods available on the platform and selects an optimal combination of forecasts in order to maximise their accuracy. Therefore, we do not have to perform large number of forecasts and then wonder which of them is the most reliable one. The Expert method is the most convenient one from all methods available on the platform. It supersedes all the previously discussed forecast methods because it always chooses the best variant.

The way how the expert method works might be divided into two stages:
  • Stage called “studying” - it verifies precision of the forecasts created by all the available forecast methods and it also assesses their accuracy.
  • Stage called “primary forecast” - which constitutes a combination of all the performed forecasts whose share in the expert forecast depends on the accuracy of respective methods.
Therefore, the expert method takes all the available forecast methods into consideration, both the optimistic ones and the pessimistic ones, so it creates a forecast that is presumably the most balanced and the most accurate one.

In order to generate an expert method there will be at least 7 observations needed.
Project co-financed by the European Regional Development Fund under the Operational Programme Innovative Economy
Grants for innovation - We invest in your future