Katkat
Back

Implement and Tune KNN Classifier

Qube Research & TechnologiesData Scientist · Online Assessment
HardData ScientistOnline AssessmentMachine Learning

Last updated

This question evaluates a candidate's competency in implementing and tuning the K-nearest neighbors algorithm, covering data preprocessing, distance-based classification, model evaluation, and hyperparameter selection.

You are given two CSV files for a three-class leaf-classification task.

  • train.csv contains three numeric feature columns: feature_0, feature_1, and feature_2, plus a class label column.
  • unlabeled.csv contains the same three feature columns but no labels.

Complete the task in a Jupyter notebook:

  1. Load and inspect the labeled dataset.
  2. Split the labeled data into training and testing sets.
  3. Implement a K-nearest neighbors classifier from scratch. You may use standard data-processing libraries, but the KNN prediction logic should be your own implementation.
  4. Train the classifier and evaluate its accuracy on the held-out test set.
  5. Tune hyperparameters, especially the value of k, and try to improve accuracy. You may use cross-validation.
  6. Use the final model to predict labels for unlabeled.csv.
  7. Write the predictions to an output CSV file.