TIM Forecasting

The TIM Forecasting add-in allows users to make forecasts for time-series data. It relies on a powerful Machine Learning engine called TIM (Tangent Information Modeler). TIM comes with a RTInstantML (Real-Time Instant Machine Learning) technology that automates everything from model building to model application, taking input data and directly providing the user with the relevant forecast. This simplifies and speeds up the whole forecasting process so it can be offered as an Excel add-in.

Users can use univariate as well as multivariate time series data with a constant sampling rate. All that is required from the user is the number of samples to be forecasted. It is that simple.

What data can be used for predictions? The Excel file must contain a date/time column with a constant sampling rate - i.e. regularly spaced data by hour, quarter-hour, month, year... The file must contain a "target" column - the column to get forecasted values for. Additionally, the file can contain columns with "predictors", i.e. data that relates to the target variable and has certain predictive power with respect to the target variables. The availability of good predictors in the dataset can improve the accuracy of the forecast.

How should the add-in be used? At first, the range of cells to be taken into account needs to be selected. The add-in tries to detect the area that contains data in the spreadsheet. If the detected range is correct, it just needs to be confirmed, if it's not correct, the user can adjust the selection of cells manually. The TIM add-in will focus on the selected range only. Afterwards, the task pane (the target and/or predictor controls) is populated with the column names that were detected. The next step is to specify how many steps need to be forecasted. Then, all that's left to do is to click the Forecast button. In a few moments the forecasted values will automatically be added to the end of the dataset range.

Version 1.1 of this add-in now contains optional extensive visualisation capabilities, including:

- a visualization of the forecast,

- predictor intervals and

- a treemap of predictor importances.

These visuaisations give users insights into the model that is used for calculating the forecast and help them to understand the importance of the different variables in their dataset.

In order to use this add-in, users need to have access to Tangent Works credentials. The registration form to get access as well as a free 14-day trial license can be opened the add-in's start screen. After the trial period is over, a paid subscription is required.

This add-in requires an internet connection to work. It calculates forecasts in the cloud from data in the spreadsheet, thus it requires the data to be uploaded - please be aware of your data plan when connected via mobile.

Learn more about the unique technology used by our add-in at https://www.tangent.works/ and https://docs.tangent.works/.