Conceptual drift detection and adaptation method based on sub-feature selection

• Main Authors: YANG LIANGHUAI, GONG WEIHUA, FAN YULEI, HUANG BUYU
• Format: Patent
• Language: Chinese; English
• Published: 28.06.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention discloses a concept drift detection and adaptation method based on sub-feature selection, and the method comprises the following steps: 1) collecting data to form a data set, constructing a classification prediction model based on an integrated classifier and a concept drift detector based on a single-class classifier, and enabling the classification prediction model and the concept drift detector to work in parallel after initialization is completed; (2) carrying out concept drift detection by using a concept drift detector, when the proportion of abnormal data points is detected to be greater than a warning threshold value, considering that drift occurs, carrying out subsequent steps and updating the concept drift detector; 3) performing sub-feature drift amplitude detection by using a binary classifier, detecting the drift amplitude of each sub-feature in the data block with concept drift, and taking the drift amplitude as an update weight of the integrated classification prediction model; an