Real tools and the ML project life cycle
We wrote everything from scratch so you understand what happens inside a model. At work nobody hand-writes k-NN: people use proven libraries. The best known:
| pandas | Read data from Excel, CSV and databases, then clean and aggregate it. Like SQL and Excel inside Python. |
| scikit-learn | The go-to library for classic ML: regression, k-NN, k-means, decision trees, all used the same way: fit then predict. |
| Jupyter / Colab | Notebooks where you write code and see results and charts right below it. Google Colab is free and runs in the browser with nothing to install. |
Here is the churn model from the k-NN lesson again, with scikit-learn. Notice how short it is. (This code needs Colab or Jupyter; it doesn't run here.)
import pandas as pd from sklearn.model_selection import train_test_split from sklearn.neighbors import KNeighborsClassifier from sklearn.metrics import precision_score, recall_score df = pd.read_csv("stores.csv") # orders, weeks, left X = df[["orders", "weeks"]] # features y = df["left"] # target X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1) model = KNeighborsClassifier(n_neighbors=3) model.fit(X_train, y_train) # learn pred = model.predict(X_test) # predict print("Precision:", precision_score(y_test, pred)) print("Recall: ", recall_score(y_test, pred))
No code? Some tools build simple models for you: Excel's Forecast Sheet forecasts monthly sales in one click, Power BI adds forecasts and anomaly detection inside charts, and cloud services such as Azure Machine Learning and Google Vertex AI offer AutoML, which tries many models and keeps the best. But the understanding you built in this track is what tells you whether a result makes sense.
The ML project life cycle: the best-known method is CRISP-DM, and it's a cycle because you often step back: