Hadoop Log Dataset for Binary Anomaly Classification

The original HDFS raw log files, which consisted of unstructured textual records, were transformed into a structured tabular CSV by extracting the essential fields of each log entry, including date, time, pid, level, component, and content. Following this conversion, the official HDFS anomaly label file, which designates log sequences as either normal or abnormal, was utilized to assign a corresponding class to each record in the tabular dataset. Through this matching process, a fully labeled dataset was produced in which every record is identified according to its normal or abnormal status, thereby enabling its use in anomaly detection and subsequent machine learning analysis.