8. Classification

Classification is the task of predicting /estimating a discrete category (the class) from input data. Training is performed using supervised machine learning, using a labeled dataset. Classification can be binary or multi-class.

Sometimes a single model has multiple outputs, multi-label classification. This is not currently supported in emlearn. Instead one classification model per output must be used.

8.1. Applications

Detecting events using Machine Learning has a wide range of applications, and it is commonly performed on sensor data using embedded systems.

Application examples

Area	Task	Sensor
Health	Detection of heart rythm irreguarity	Electrocardiogram (ECG)
Wearables	Scene classification for contextual awareness	Sound
Farming	Cattle behavior classification for health tracking	Accelerometer
Robotics	Material identification for grippers	Capacitive
Buildings	Human presence detection	Radar

8.2. Classification models

Supported classification models

Algorithm	Implementation
RandomForest	RandomForestClassifier
ExtraTrees	ExtraTreesClassifier
DecisionTree	DecisionTreeClassifier
Multi-Layer-Perceptron	MLPClassifier, keras.Sequential
Gaussian Naive Bayes	GaussianNB
Nearest Neighbors	KNeighborsClassifier

A basic example showing some of these classifier models can be found in Classifier comparison.

8.3. Related

Classification is an important component of Event Detection systems.