If most of the K most similar samples of a sample in the feature space belong to a certain category, then the sample also belongs to this category.

API

distance

The k value is too small: easily affected by outliers - overfitting

The k value is too large: it suffers from the problem of sample balancing-underfitting

Approximation error (focusing on the training set) - overfitting - performs well on the training set, but performs poorly on the test set

Estimation error (focusing on the test set) - smaller, indicating better prediction of unknown data

Construction of kd tree

nearest neighbor search

Obtaining data sklearn.datasets

Introduction to data set return values: The return value is batch (dictionary)

Data visualization: seaborn(x=,y=,data=,hue="target value",fit_reg=False does not perform data fitting)

Data set division: sklearn.model_selection.train_test_split (x=feature value, y=target value, test_size=0.2 test set size, random_size=22 random number seed)

feature engineering

Simple and effective, low retraining cost, suitable for cross-class samples, and suitable for automatic classification of large samples

Lazy learning, category scoring is not standardized, is not good at unbalanced samples, and requires a large amount of calculation

Get dataset

Basic data processing

feature engineering

machine learning

Model evaluation