Forecasting Methodologies

Challenges

 

The understanding and knowledge of the future are essential to good planning and organization. Off the cuff, you need to have tools that perform reliable forecasts on key business variables.

 


Approach

 

Forecasting methods allow one to predict the behaviour of the same element in the future, based on past behaviour and business environment conditions.
Depending on the business environment, the future behaviour of an element can be governed by stationary or work caseloads, or dispose a non-periodic behaviour.

Depending on the casuistry of business and customer needs, AIA has different forecasting methods to suit the behaviour of information to provide.
Symbolic Forecasting

This forecasting method is an AIA proprietary design and is an improved combination of techniques:
– Seasonality Decomposition
– Exponential Smoothing

The seasonality decomposition allows one to naturally reflect the periodic components of the series. This decomposition is complemented with exceptionalities treatment allowing for situations that are out of preset cycles.

The symbolic foresight is very suitable to model behaviours hardly vested, such as the treatment of special days. Its features allow making forecasts with little or no history, and proper complex behaviour modelling.
Finally, it should be noted that symbolic foresight is extremely fast because it needs little information to forecast. To calculate new forecasts many algorithms need to recalculate calculate it from all historical, while Symbolic Forecasting requires only the latest update of a few parameters, which is especially suited to applications with a high volume of forecasts or systems that perform real-time forecasting.

 

 

Aggregate Demand Forecast

Aggregate Demand characterizes provident solutions:

  • User expert in business and technology foresight.
  • The data quality is excellent.
  • The quality forecast requirements are very high.
  • Reduced data volume. Office environment.
  • Small number series (between 1 and 10).

Grupo AIA provides solutions for this type of forecast, but foregrounds
the product of aggregate demand Forecast Power, which has been specialized
for forecasting electricity demand time (active or reactive)
for aggregate consumption of a group of consumers (a region or an area).

The product Aggregated Demand Electricity Forecasting offers the following
features for planning and operations teams:

  • Hourly demand active and reactive energy.
  • Daily, weekly and monthly reporting of the analysis.
  • Interconnection in real time (optional).
  • Tools that allow users to enter and modify the values: hourly values, definition of special days and temperature values.
  • Analysis of historical series, which includes temperature, season, and day of the week, for forecasting and to generate the characteristics of the profile shape.

 

Distributed Demand Forecasting

Distributed Demand Forecasting solutions are characterized by:

  • User expert business, but do not need to know the foresight technology.
  • Variable data quality. It requires a historical reconstruction in the solution.
  • Quality estimates are required, but to a reasonable cost.
  • Huge data volume. Client-server architecture to 3-tier (web)
  • High number series (between 500 and 5000).

Grupo AIA provides solutions for this kind of foresight.

 


Benefits

 

AIA forecasting methods allow:

  •   Operational optimisation based on detailed knowledge of future behaviour.
  •   Development of strategies for short and long term
  •   Estimation of the impact of changes in business variables and trends.
  •   By modelling the whole problem, the algorithms developed by AIA enable:
    • What-if simulations
    • Trend analysis
    • Sensitivity of information to business variables.

 


Success Cases

 

AIA forecast algorithms have been successfully applied in different environments:

  • Forecasting electricity and gas consumption.
  • Forecasting electricity market diversion.
  • Forecasting Switching customers.
  • Forecasting equipment breakdowns in electrical nodes.
  • Forecasting cash demand in ATMs and branches.
  • Forecasting telecommunications traffic.
  • Forecasting electricity market prices.