Identifying and Modeling Families of Serial Tests
Ciena is a leading corporation supplying IT products. Before delivering any product, the company makes sure all products go through a set of tests which are recorded into log files. Operating in a crowded and highly competitive market, Ciena is continuously running after innovation for remaining highly competitive. Therefore, the company wants to increase their data analytical solutions capabilities by exploring the huge amount of log data that are continuously gathered. The diversity and the time evolving aspect of the generated log information yield complex and massive data which are difficult to handle. Digging into such data pose new challenges in elaborating effective mining algorithms. The aim of this project is to devise analytical tools to fully grab the massive heterogeneous generated data. To this end, we intend developing novel datamining / machine learning models that extract useful patterns that offer insight to track the status of products tests over different time stamps. The goal is to infer the hidden inter-dependencies between test cases and thus build a dynamic multidimensional heterogeneous network structures for mimicking the logs ecosystem environment to first identify families of tests. Then, model these families to finally predict early symptoms of failure products.