Basic Model Builder

The Basic Model Builder is ideal for users seeking a quick and efficient way to create models with minimal configuration. It leverages sensible defaults to automate much of the process, allowing you to focus on high-level decisions while the platform handles the underlying complexities. To get started, simply click the New Model button, make a few adjustments, and let Alviss AI build the model automatically.

This builder is particularly useful for initial explorations or when you need rapid insights into how various drivers (e.g., marketing, pricing, or seasonal trends) impact your key performance indicators (KPIs), such as sales or churn.

Steps to Build a Model

1. Select Dataset
   Choose the Dataset that will serve as the foundation for your model. This dataset provides the raw data for analysis, including variables like sales figures, marketing spend, or external factors. Ensure the selected dataset aligns with your project's periodicity and has been validated for quality.

2. Choose Modeling Combination
   Specify the [modeling combinations](Modeling Combination.md) based on dimensions such as country, region, or grouping. For each unique combination (e.g., "US - East Coast - Retail"), Alviss AI will build a dedicated model. This allows for granular, tailored insights while maintaining scalability across your data.

3. Set Effects
   Define the variable groups to include in the model, drawn from your dataset. Common groups include Media (e.g., ad spend), Season (e.g., holiday effects), Trend (e.g., long-term market shifts), Pricing, and more.

   • Optionally, indicate the expected directional impact of each group on your target KPI (e.g., positive for marketing investments increasing sales).
   • Note: Including more groups enhances model comprehensiveness but may increase training time due to added complexity.

4. Select a Date Range
   Define the time periods for model training:

   • Training Period: The historical data range used to fit the model.
   • Hold-Out Period: A separate range reserved for validation, helping assess the model's predictive accuracy on unseen data.

5. Advanced Settings (Optional)
   Customize technical parameters for finer control:

   • Adjust the number of samples, learning rate, and epochs to influence training efficiency and model convergence.
   • Enable post-training actions, such as generating Predizctions, Attributions, or comparing Prior vs. Posterior distributions for deeper insights.
   • Opt to automatically [activate the model](Active Model.md) upon completion, making it immediately available for use in dashboards like Effect or Activities.

Once submitted, Alviss AI will train the model in the background. You'll receive notifications upon completion, and the model will appear in your project's model list for review, activation, or further refinement.