Step 3. Sales and demand forecasting: Algorithm selection

PostedFebruary 22, 2022

UpdatedMarch 14, 2022

ByEdil Tilekov

Algorithm Selection

Use Algorithm selection screen to select the algorithm and parameters (if applicable). Selection can be done only for licensed algorithms. User can create multiple projects for the same algorithm to support multiple output requirements. However, single algorithm only can be selected per one project.

Note: Algorithm parameters cannot be changed when the user makes a selection and moves to the following screen.

Algorithm selection – from 1:44 to 3:55

Inventory Demand Forecasting

This algorithm is using the Sales Forecasting algorithm as a base for predictions and it is allowing to users to predict how many products will be sold in the selected period. Actionable outputs of this algorithm are indicating what is the current level of your stock and if you need to purchase more items to satisfy future customer demand or if you are way over with stored on-hand quantities. This algorithm is running solely on a local computer or server and none of the data is stored on the cloud.

Notes:

Artificial intelligence is used to recommend relevant reorder quantities however users are still in full control of procurement or manufacturing activities.

Units of Time for Training

This drop-down list is indicating what will be data consolidation period. For example, when the user selects ‘Month’ input sales data for training will be consolidated per customer, product and month even if the daily sales CSV file has been imported. It is recommended to keep ‘months’ for monthly predictions and ‘days’ for daily and weekly predictions.

Units of Time for Forecasting

This drop-down list is indicating what is time period you want to generate a forecast for. Available units of time options for forecasting are Days, Weeks, Months.

No. of rounding decimals

This drop-down list is indicating how many decimals you will use on the results of the algorithm and it is strongly dependent on what type of products you are selling. Most common choice for the majority of customers is 0. User needs to create a new project if a demand forecast should be generated for products with various decimal places.

Forecast length

This value is indicating for how many days, weeks or months you want to calculate your forecast. There is no limitation on forecast length.

Note: Keep in mind that selection of long forecasting period length might result in decreased accuracy of forecast at the end of the selected period as the system is utilising generated predictions as data input.

Advanced options

This is the place where user can select external indicators to be used during the training of the algorithm. This section is under development (status as of 19th June 2021)