←Tutte le torri

Raahaao Code: Machine Learning

Teach computers to learn from data, and know when not to trust them.

9+ ore di lezioni · 2 progetti · €35

Ti aiuta a prepararti per: AWS Certified AI Practitioner, AIF-C01 (AWS) · Azure AI Fundamentals, AI-901 (Microsoft)

Oppure includila in un piano mensile da €3.46 a settimana (addebitati €15 al mese) →

Accedi per iniziare

Serve un account per aprire qualsiasi lezione. Senza account puoi sfogliare le torri e i prezzi.

Paga in

Cosa imparerai, piano per piano

  1. 1

    What learning means (Beginner)

    • What machine learning is
    • Your first model
    • Train and test
  2. 2

    Supervised learning (Beginner)

    • Classification
    • Measuring a classifier
    • k-nearest neighbours and scaling
  3. 3

    Better models (Intermediate)

    • Decision trees
    • Ensembles and random forests
    • Cross-validation and tuning
  4. 4

    Beyond labels (Intermediate)

    • Clustering with k-means
    • Dimensionality reduction with PCA
    • Working with text
  5. 5

    Pro (Job-ready)

    • Neural networks from scratch
    • Feature engineering and pipelines
    • Fairness, bias and ethics
    • From notebook to product

I due progetti sul tetto

House price predictor

Build a small house-price prediction workflow from a supplied CSV. Prepare input features without exposing the sale price to the model, train a regression pipeline with a reproducible holdout split, and measure its errors.

  • load(path) reads the house-price CSV and returns a pandas DataFrame
  • make_features(df) returns the four input columns and never includes price_eur
  • train(df) returns a fitted sklearn Pipeline trained on a deterministic training split, not the held-out rows
  • evaluate(model, df) returns a dict with numeric mae and r2 values for the supplied rows
  • predict_one(model, house) returns a float prediction for one valid house
  • predict_one(model, house) raises ValueError for missing or non-finite/non-numeric feature values
  • Training and evaluation do not use the target as an input feature, and training is deterministic

Churn early-warning

Build an early-warning classifier for customer churn using the supplied, imbalanced customer data. Train a reproducible pipeline that keeps the churn label out of its inputs, then measure errors that matter for finding customers at risk.

  • load(path) reads the churn CSV and returns a pandas DataFrame
  • train(df) returns a fitted sklearn Pipeline using only the four declared input features and a deterministic stratified train/test split
  • train(df) fits on the training partition only and accounts for the imbalanced churn classes
  • evaluate(model, df) returns precision, recall, and a 2x2 confusion matrix for the supplied rows at threshold 0.5
  • choose_threshold(model, df, min_recall) returns the highest probability threshold that reaches the requested recall
  • explain(model) returns the top three original input features ranked by summed absolute model coefficient
  • fairness_report(model, df, group_col) reports sample size, precision, and recall for every group

Domande

Devo installare qualcosa?

No. Tutto funziona nel tuo browser. Per i progetti puoi anche lavorare sul tuo computer e caricare i tuoi file.

Posso saltare quello che conosco già?

Sì. Una volta che possiedi una torre, ogni piano ha un test di salto: segna 8 su 10 e quel piano si apre.

Qual è la politica di restituzione?

Rimborsi entro 5 giorni dall'acquisto. Dopo 5 giorni, nessun rimborso o reso.

Mantengo l'accesso?

Sì. Una torre che acquisti è tua, senza abbonamento.

Refund policy

Le torri

HTML & CSS · JavaScript · Python · SQL · C++ · Node.js · Python per i dati · Flask · Logic & Problem Solving · Algorithms & Data Structures · Java · Building with AI