A Solution Template describes how to apply TIM to a specific business problem. The template consists of the following sections:
- Problem description,
- Data recommendation template,
- Demo example.
Each of these sections will be explained in more detail below.
This section describes a specific problem and its relationship to forecasting or anomaly detection. A problem description typically consists of a business problem explanation, a list of associated KPI’s and a common forecasting routine.
Data recommendation template¶
Garbage in, garbage out (GIGO) is the well-known concept that the quality of a system’s output can only be as high as the quality of its input. In this section, a list of recommended explanatory variables is provided. No matter the quality of the modeling technique, without appropriate explanatory variables the resulting model will not achieve a satisfactory performance.
A tempting remedy is to increase model complexity and try again. This is however not a real solution, as it often results in over-fitting of the data. To give an idea of what results can be expected, TIM provides a ‘data difficulty’ measure that indicates how well the provided explanatory variables are likely to explain the variance of the target variable.
Most of the time, TIM requires no setup of mathematical internals and produces excellent results with default settings. A user only must provide a forecasting routine and a desired prediction horizon.
This section consists of a step-by-step walkthrough explaining how to build a TIM Solution, including a demo dataset.