Disaster relief operation is refer to an activity where people assist the disaster victim to recover. Inefficiency of distribution centre selection in disaster relief operation makes difficulty for volunteer to perform their humanitarian task. Thus, a strategic location choose of operation centre is being a concern. It has been pointed out that not all disaster area can be covered during disaster recovery operation . The problem of the selection of distribution centre is not done optimally. The methodology by comparison between K-means, K-means with Simulated Annealing (SA), lastly K-means with Genetic Algorithm (GA). In response to the problems, it is needed to understand the existing algorithm used to determine the distribution centre. The K-Nearest Neighbor (KNN) and the use of Genetic Algorithm (GA) and Simulated Annealing (SA) in KNN is proposed to classify and select the distribution centre. The minimization of the fitness value is being the objective in this study. The experiment conducted with demand point and the distribution centre are located by researcher and complement the KNN with the GA and SA. The comparison of the performance has been made and the study found that implementing GA-KNN give the most optimal solution with average 21% of fitness value least compared to SA-KNN. Thus, this study is contributing in finding the most optimal distribution centre location with nearly-equal demand point distribution of each selected location which can facilitates the real-world aid distribution in disaster area, time-wise and cost-wise.

In the telerehabilitation system, the statistical data of the patients’ movement are stored in the temporary storage and synchronised to the storage service of online cloud data. Application providers faced a problem in reducing the monetary cost of the whole cloud service and reducing the footprint of the main memory space. In addition, users encounter long latency when the required data need to be read from the cloud via the internet and the hard disk drive (HDD) of the cloud servers. To solve this problem, an optimal data access framework is presented to cache the statistical data of the patients in the application server. The main memory database and cache use internal tracking in the main memory to track records that are not accessed by transferring the data to the disk. This mechanism retains the keys and all indexed fields of evicted records in the main memory which prevents potential memory space savings for the application that have many keys and secondary indexes. Therefore, to overcome the mentioned problems, the cloud database is categorised into three partitions (hot, warm, cold). In addition, a cache memory image in the application server is provided for the hot partition of the cloud database. The use of cache memory image reduces the number of reading operations from the cloud and saves the space of the main memory. The experimental results showed that the proposed framework can produce good quality solutions by utilising the main memory space and reducing the latency and read operations from the cloud that lead to reducing the monetary costs.