Preprocess, Train and Evaluate a Model | Dataford Interview Questions

Given a dataset, write the preprocessing and cleaning code, train a simple model in a framework of your choice, and report evaluation metrics (craft exercise)

Asked in the craft exercise stage. This is a multi-part, two-page prompt in CoderPad. Emphasise presentation of results and next steps.

Implement train_model(dataset, target_key). Each dataset row is a dictionary containing numeric feature values, with missing values represented by null, and a binary target. Remove exact duplicate rows, discard rows with a missing target, impute missing numeric features using training-set means, standardize features, and train a nearest-centroid classifier. Use a deterministic stratified holdout containing the last row from each class. Return accuracy, precision, recall, and F1, rounded to four decimal places.

Constraints

Each row is a dictionary with at least one numeric feature.
The target is 0 or 1 when present.
Each class has at least two valid rows after cleaning.
Missing feature values are represented by null.
Use only the Python standard library.

Function Signature

def train_model(dataset, target_key):

Your question is Preprocess, Train and Evaluate a Model. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem

Given a dataset, write the preprocessing and cleaning code, train a simple model in a framework of your choice, and report evaluation metrics (craft exercise)

Asked in the craft exercise stage. This is a multi-part, two-page prompt in CoderPad. Emphasise presentation of results and next steps.

Constraints

Each row is a dictionary with at least one numeric feature.
The target is 0 or 1 when present.
Each class has at least two valid rows after cleaning.
Missing feature values are represented by null.
Use only the Python standard library.

Function Signature

def train_model(dataset, target_key):

Your question is Preprocess, Train and Evaluate a Model. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.