Amount Of Data Access Research Articles

Chameleon: A Self-adaptive cache strategy under the ever-changing access frequency in edge network

In recent years, with the maturity of 5G and Internet of Things technologies, the number of mobile applications and the amount of data access have increased explosively. However, the frequency of these accesses varies considerably at different times of the day, requiring different caching strategies in those limited-capacity edge servers. Existing caching strategies perform well when the access frequency is stable. However, they ignore the time-varying characteristics of user access frequency in different periods, resulting in a low hit rate in ever-changing frequency scenarios. To improve the hit rate in such scenarios, we propose a cache replacement policy called Chameleon, which consists of two components, AutoFre, and Crates. AutoFre is an admission algorithm that predicts the future access frequency category and calculates the admission thresholds based on the prediction result. While Crates is an eviction algorithm, it selects the contents evicted by designing a customized principal component analysis algorithm. We conduct a series of experiments with real application traces from ChuangCache. The trace has 9,839,213 user accesses in 48 h. The results demonstrate that Chameleon reaches about 98% in caching hit rate and outperforms SecondHit-Crates algorithm about 8% in frequency-changing edge networks.

Cost Efficient Query Optimization in Mobile Environment

Today, we live in the world of internet. With the advancement of technology, the amount of data access has increased too many folds. Internet access now is not only limited to computer devices but can now be easily accessed through mobile devices viz. Smartphones, tablets, PDA’s. The internet is now available to every common man, and with its use he fires many queries on servers and uploads or downloads data from the internet. In fact, 90% of the world’s data came in existence in the last three-four years, and that too because the internet is readily available to each and every common individual. Of these, much data is being uploaded and queried upon by mobile devices. As the number of devices for Internet access has increased, and so is the number of queries fired by the users on a particular server. The time taken by a query to process totally depends on the complexity involved in joining the tables distributed along the network and finally extracting the desired result out of it. Processing and optimization of various queries in mobile devices involve much join computation among data present at different sites that may be static or mobile which in turn requires much energy consumption. A mobile device has limited energy, so, it must be utilized efficiently. Much research work have been done till now, in the field of mobile computation and making efficient use of energy. However, as the mobile devices possess some asymmetric features, and because of that the old techniques used in distributed databases cannot be applied directly. This paper brings out some methods, to efficiently utilize mobile energy by employing per split semi-join using MapReduce Framework of Hadoop. General Terms Per-Split Semi-Join, MapReduce, Hadoop, Cost Optimization, Distributed Databases

Improved Data Mining Algorithms Based on an Early Warning System of College Students

In order to solve the problem of early warning of college students’ achievement, this paper proposes two improved algorithms for data pre-processing and mining warning factor. At first we put forward an improved K-Means algorithm which is not only ensures the accuracy of the original algorithm, but also improves the stability of the algorithm. Then we put forward an improved algorithm New_Apriori algorithm and analyze experiment result. The result shows that the amount of data access has been reduced significantly and efficiency has been improved. In the end of this paper, we built the early warning model of students’ achievement based on neural network. The result of experiment shows that the new algorithms improve the efficiency and accuracy of the early warning.

Parallel frequent set counting

Computing the frequent subsets of large multi-attribute data is a key component of local pattern detection data mining algorithms. It is both computation- and data-intensive. The standard parallel algorithms require multiple passes through the data. The cost of data access may easily outweigh any performance gained by parallelizing the computational part. We address two opportunities for performance improvement: using a parallel approximate algorithm that requires only a single pass over the data; and using a probabilistic technique to avoid generating most of the lattice of subsets implied by each object's data. The computation required is only slightly greater than levelwise algorithms, but the amount of data access is much smaller.