Contact us
Call us if you want to find out how we can help your organization :
+7 495 648 69 18 (Moscow) Russian speakers
Weekdays, 9AM - 6PM, GMT + 3
+420 773 052 550 (Prague) English speakers
Weekdays, 10AM - 7PM, GTM + 1

Time series forecasting

Success in reaching the target goals depends mainly on how much time it takes to submit the administration and responsible executives a precise and comprehensive forecast of further course of events. Complex business processes, such as retail trade, financial markets monitoring, logistic code planning, etc. in the course of their activity accumulate large databases of time-ordered observations. These data can be presented as time series. Time series forecasting is one of the most challenging contemporary tasks that is inextricably connected with optimum decisions making and enterprise efficiency management.

Advanced analytics is applied to solve tasks where it is necessary to forecast in a speedy fashion a lot of time series (from thousands up to hundred thousands). These can be financial instruments quotetions, product range, web-sites traffic, demand for goods and services, etc.

A wide range of special methods – automatic accounting of trends, seasonality, autocorrelation, time series groups interrelation, external events (news, marketing campaigns, promotion actions, etc.) – is used to provide the most precise forecasts. There is also a possibility to create hierarchy forecasts (e.g., hierarchy forecasting is required in conditions of nomenclature regular changes in retail trade).

Faunus Analytics provides a full range of services for data preparation and time series forecasting. We help to reduce the uncertainty level that is faced by any business, providing our clients with reliable answers to the question how the events will develop in future. We offer the following:

  • Building of time series history on the basis of transaction data
  • Analysis of such events as financial and political news, holidays, week-ends, promotion actions, marketing campaigns, outstocks, varying macroeconomic indicators, industry indices, etc.
  • Data preprocessing to reduce noise level and recognize events important for the forecast
  • Analysis of time series general properties, such as stationarity, ergodicity, heteroscedasticity, presence of cycles, trends, autocorrelation properties, etc.
  • Time series grouping
  • Determination of the structure and estimation of statistical models parameters
  • Estimation of statistical models accuracy and their selection
  • Automatic detection of time series requiring additional expert analysis
  • Forecasting on the basis of the created statistical models
  • Monitoring of the forecasting process efficiency as a whole.

See also