How to Conduct Machine Learning
Use machine learning when you want to identify which predictors have the greatest impact on the outcome.
Step 1: Select Your Machine Learning Task
- Click Machine Learning in the sidebar menu.
- Choose the type of task:
- Regression — predict a numeric outcome (e.g., Revenue).
- Classification — predict a category (e.g., Churn: yes/no).
Step 2: Configure Your Model
- Choose the outcome variable (target you want to predict, e.g., Revenue).
- Choose predictor variables (features used for prediction, e.g., satisfaction, brand loyalty, burnout, address mismatch).
- Click Apply model settings.
Step 3: Review Model Results
- Examine key metrics (e.g., R² = 0.56 for regression).
- Check feature importance to see which variables most influence predictions.
- Example: Product satisfaction is a strong positive driver of revenue.
- Interpret the model:
- Understand how changes in predictors affect the outcome.
- Identify actionable insights (e.g., improving satisfaction may increase revenue).
Step 4: Report and Act on Insights
- Look for Averisys AI and click Results to view the summarized findings:
- Clearly state which predictors most influence the outcome.
- Highlight any practical implications (e.g., customers with low satisfaction are more likely to churn).
- Develop action plans based on insights (e.g., target customers at risk of churn, optimize pricing).
- Click Reports to see the executive summary.
- Click Publish this page to share results with stakeholders via a published page.
- Click Invite Users to invite stakeholders to share results via dashboards.
Tips & Best Practices
- Clean and preprocess your data for the best results.
- Use feature importance to guide business decisions.
- Regularly update and retrain your models with fresh data.
If something does not work as described
Screens are updated as the platform changes, so a button may sit in a slightly different place than the wording above suggests. When that happens, the AI assistant inside the platform is the fastest way forward: it can see the page being worked on and answer in context, at any hour. Ask it what a message means, why a file was rejected, or which step comes next.
Most problems in this part of the workflow come from the data rather than the platform. Column headers that repeat, numbers stored as text, blank rows at the bottom of a file, and inconsistent category labels are the usual causes. Checking those four things resolves the majority of failed uploads and unexpected results before any further troubleshooting is needed.
For anything that still does not resolve, email support@averisysanalytics.com with the guide title, the step reached, and what appeared instead of the expected result. Including the file name and a screenshot usually means the issue is answered in the first reply. Related instructions are listed in the full platform guide, and common questions are collected in the support FAQ.