Generate an Idea for Model Training
Start by defining the specific goal the model should predict, such as customer churn, likelihood of purchase, or response to a marketing campaign.
For inspiration, explore ideas in our ML portfolio and case descriptions to identify use cases aligned with your business objectives.
Create Historical Datasets for training
- Metrics Dataset: Compile comprehensive historical data on the objects of prediction (e.g., customers), covering behavioral, demographic, and transactional metrics. The quality of your training depends on the breadth and depth of your dataset, covering all measurable interactions between customers and your company across various channels. Examples include purchase counts, discounts redeemed, support interactions, app/web activity, communication outcomes, product preferences, etc.
How to Prepare Training Dataset with Metrics
- Targets Dataset: Assemble historical data that directly relates to the business objective, representing measurable outcomes (targets) such as customer churn (0/1) or purchase frequency. Ensure targets are well-defined, time-bound, and actionable. Experiment with different target types—binary, multiclass, or regression—to achieve the most accurate and relevant results for your needs.
How to Prepare Training Dataset with Targets
Upload Datasets to Bind Genius
Load both the metrics and targets datasets into the system. Use tools like CSV files for quick experiments and PoC or BigQuery for automated data integration.
How to Add Dataset
Add a new predictive ML Model
Configure the new model by choosing a name, type (binary, multi-class, or regression), and other training parameters.
How to Add and Train ML Model
Evaluate Model Quality
Compare the model's performance against your business requirements and determine if the model is suitable for deployment or requires further refinement.
If results are unsatisfactory, adjust data, parameters, or features and retrain the model.
How to Evaluate ML Model (binary and multi classification)
How to Evaluate ML Model (regression)
How to ReTrain ML Model