Figure 1.

Principles of ECG restitution analysis using k-nearest neighbors classifier machine learning technique. (A) Schematic of data classification using the k-NN classifier with k=3. Data points (crosses) are classified as either in Class 1 or Class 2 based on their k closest neighbors (a,c,d—Class 1, b,e,f,g—Class 2). (B) Data flow for the patient classification using the k-NN classifier. Extensive preclassified training set (red and blue dots) is used to create the model. Query points from the patient’s records (crosses) are classified individually by classification of three neared neighbors and the outcome for each is decided by a simple majority vote. The final outcome is then decided by subsequent majority vote on all records for that patient.