| Popis: |
Computer systems are becoming increasingly complex, they are growing dynamically due to the mobility of their devices, multi-operating environments, frequent updates and upgrades. Conventional reliability theory and conventional methods rarely take into account the actual state of the system; thus, they are not able to reflect the dynamics of the execution systems and the failure processes. The variety of models used illustrate the lack of consensus on how to apprehend this phenomenon, especially the difficulties to formulate a general model of interpretation. Moreover, this diversity of approaches has been developed mainly to diagnosis computer systems. Consequently, the nature of the information available on these systems and the type of defects to be detected lead to the implementation of specific strategies. The main objective of this paper is to use an Iterative Partitioning Log Mining algorithm (IPLOM) to cluster logs events before they are processed further by an optimized deep learning model to detect anomalies form large amount of IBM BlueGene/L collected data. ALL the features are classified by the score and also selected to be saved or even removed from the dataset. The procedures are frequently univariate and consider the feature independently, or with regard to the dependent variable. In such methods, the subset selection process is independent of the learning algorithm, and it is a pre-processing step. Generally, this work describes how to optimize deep learning models in different types of machine learning problems. We started by developing a mining model, and then apply a filter to the input data in order to eliminate the correlated information. Then, we compared the results of these different models using an elevation curves graph. Finally, we extracted additional knowledge from the underlying mining structure. This leads to optimize learning features and to reduce the classification time and even improve the performance of systems’ rules, particularly in resolving complex problems involving a large number of input data. |
