Correlation and root cause analysis of trace data using an unsupervised autoencoder
Abstract:
An analyzer system inputs parameter values from trace files of a software application into an autoencoder. The analyzer system adjusts weights of the edges between nodes in the autoencoder until reconstruction errors in outputs are minimized. The analyzer system receives a selection of a parameter represented in an autoencoder. In response, the analyzer system identifies hidden layer nodes connected to an output node corresponding to the selected parameter and identifies other output nodes connected to the hidden layer nodes. The analyzer system retrieves weights assigned to edges between the hidden layer nodes and the other output nodes. The analyzer system calculates correlation values between the output node corresponding to the selected parameter and each of the other output nodes and outputs the correlation values. A user can use the correlation values to better direct the root cause analysis.
Information query
Patent Agency Ranking
0/0