Analysis #2

• medium 7.24% of all cells are missing (398 cells).
• low 1 identifier-like column(s) detected and excluded from modelling: Name.
• medium High-cardinality categorical column(s) that will be grouped before encoding: Email.

The target is continuous with 449 distinct numeric values.

R² reports the share of variation in the target the model explains, which is comparable across datasets in a way that RMSE is not.

Random 80/20 split with random_state=42.

Strongest signal against the target

The AI layer suggested model names; the agent resolved each one against a fixed library of hand-written models and refused anything that did not fit.

Suggestions the agent refused

• Logistic Regression — It is a classification model, but this is a regression problem.
• Decision Tree Classifier — It is a classification model, but this is a regression problem.
• Random Forest Classifier — It is a classification model, but this is a regression problem.
• Extra Trees Classifier — It is a classification model, but this is a regression problem.
• K-Nearest Neighbors Classifier — It is a classification model, but this is a regression problem.

Ridge Regression achieved the best R2 of 0.0462, ahead of Linear Regression (0.0432) by 0.0030. R2 was chosen because it matches what the objective is asking for.

Linear regression with a penalty that keeps correlated features stable.

The size of each feature's coefficient after scaling. Larger means a one-unit change in that feature moves the prediction more.

• Cleaning removed 0 row(s) and 1 column(s) that could not contribute to a reliable model.
• The winning model's decisions are driven mainly by Currency_GBP (13%), LoyaltyPoints_minus_CustomerID (12%), SubscriptionType_Standard (10%).
• On average the prediction is off by 299.59 in the target's own units, which is the number to quote when someone asks how accurate it is.
• The objective was partially met. The model explains only 4.6% of the variation, which is too little to be useful. Success criterion: A model that clearly beats the naive baseline on F1 and behaves sensibly on the outcome the objective cares about.

1. Collect more data. 500 rows is a small sample, and the score above could move considerably on a larger one.
2. Look for missing explanatory variables. When R² is low, the cause is usually that the drivers of the target were never recorded, not that the model is wrong.
3. Tune Ridge Regression with a grid or randomised search. This run used sensible defaults with no hyperparameter search, so there is headroom left.
4. Review the 4 removed feature(s) with a domain expert. Automated selection is statistical; it does not know which columns matter to the business.
5. Before deploying, re-test on data from a later time period than the training data. A random split cannot tell you whether the pattern holds next quarter.