Time series prediction using adaptive association rules

Grid computing is formed by a large collection of interconnected heterogeneous and distributed system. One of the grid computing purposes is to share computational resources. The efficiency and effectiveness of resource utilization of a grid greatly depend on the scheduler algorithm. The scheduler is able to manage the grid resources more effectively if we able to predict and provide it with the future state of grid resources. Therefore, this paper proposes a model to perform time series prediction using adaptive association rules. This model uses the idea that if a segment of a repeatable time series pattern has occurred, it has the possibility that the following segments of the repeatable pattern appear. Data mining and pattern matching techniques are being applied to mine for repeatable time series patterns. This model has the ability to provide confident level for each prediction it made and perform continuous adaptation. A prototype of this model is being developed and tested with four test cases. These test cases are relatively simple because our work on this time series prediction using adaptive association rules is very much in its early stages. The result from the experiment shows that our model is able to capture repetitive time series patterns and perform prediction using those patterns. However, this model has some drawbacks such as it required high computational power and required large storage.