Behavior sequence anomaly detection method and system based on unsupervised algorithm

• Main Authors: YIN QIAN'AN, XU MING, TAO JINGLONG, LIU SHENG, WANG QIFAN, YU XIANZHE, LIANG SHUYUN, WEI GUOFU
• Format: Patent
• Language: Chinese; English
• Published: 30.05.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention provides a behavior sequence anomaly detection method and system based on an unsupervised algorithm, and the method comprises the steps: carrying out the standardization processing of operation data of an enterprise web system into a web user operation table; segmenting the user behavior sequence at equal time intervals according to the sequence of user operations to form subsequences of a plurality of time windows; according to a web user operation table, historical data of recent four days are obtained, and for each user every day, operation behaviors in one day are combined according to an operation sequence to generate a sequence; calculating the similarity between each sub-sequence and the sub-sequences of the adjacent windows before and after the sub-sequence and the similarity between each behavior sequence and the behavior sequences in the next three days; and taking the similarity values between the subsequences and the similarity values between the sequences as features, namely input o