Lokesh Kumar Sharma

Nearest Neighbour Classification for Trajectory Data

Trajectory data mining is an emerging area of research, having a large variety of applications. This paper proposes a nearest neighbour based trajectory data as two-step process. Extensive experiments were conducted using real datasets of moving vehicles in Milan (Italy). In our method first, we build a classifier from the pre-processed 03 days training trajectory data and then we classify 04 days test trajectory data using class label. The resultant figure shows the our experimental investigation yields output as classified test trajectories, significant in terms of correctly classified success rate being 98.2. To measure the agreement between predicted and observed categorization of the dataset is carried out using Kappa statistics.

Connectivity and Coverage Preserving Schemes for Surveillance Applications in WSN

In applications such as surveillance and target monitoring, high degree of coverage and connectivity are required. This paper investigates the problem of energy efficient coverage and connectivity for random placement of nodes such that active sensor nodes are minimized. We introduce an algorithm based on connected dominating set (CDS) and use it as a virtual backbone for network connectivity. Some nodes are refined from isolation to the backbone network, while others are connected under the tributaries of backbone network. If all the nodes are activated simultaneously, it leads to redundancy and wastage of resources in the network. In our work, coverage is achieved such that overlapping area is minimized, while connectivity of network is maintained via backbone network and its tributaries.

Entropy Weighting Genetic k-Means Algorithm for Subspace Clustering

This paper presents a genetic k-means algorithm for clustering high dimensional objects in subspaces. High dimensional data faces data sparsity problem. In this algorithm, we present the genetic k-means clustering process to calculate a weight for each dimension in each cluster and use the weight values to identify the subsets of important dimensions that categorize different clusters. This is achieved by including the weight entropy in the objective function that is minimized in the k-means clustering process. Further, the use of genetic algorithm ensure for converge to the global optimum. The experiments on UCI data has reported that this algorithm can generate better clustering results than other subspace clustering algorithms. General Terms Data Mining

Efficient Trajectory Pattern Mining for both Sparse and Dense Dataset

comprehension of phenomena related to movement - not only of people and vehicles but also of animals and other moving objects - has always been a key issue in many areas of scientific investigation or social analysis. Many applications track the movement of mobile objects, using location- acquisition technologies such as Global Positioning System (GPS), Global System for Mobile Communications (GSM) etc., and it can be represented as sequences of time stamped locations. In this paper, we analyze the trajectories of moving vehicles and we develop an algorithm for mining the frequent patterns of Trajectory data. We use the extensions of sequential pattern mining to spatiotemporal annotated sequential patterns. The description of frequent behaviors in terms of both space (i.e., the regions of space visited during movements) and time (ie, the duration of movements). In this paper an efficient trajectory pattern mining is proposed by incorporating three key techniques. In this paper we have examined ways of partitioning data for trajectory pattern discovery. Our aim has been to identify methods that will enable efficient counting of frequent sets in cases where the data is much too large to be contained in primary memory, and also where the density of the data means that the number of candidates to be considered becomes very large. Our starting point was a method which makes use of an initial preprocessing of the data into a tree structure (the P-tree) which incorporates a partial counting of support totals

Associative Classifiers for Predictive Analytics: Comparative Performance Study

A new predictive modelling approach known as associative classification, integrating association mining and classification into single system is being discussed as a better alternative for predictive analytics. Our paper investigates the performance issues of significant associative classifiers likes CMAR and CPAR. Performance comparisons observe that CPAR achieves improved performance as compared to CMAR. We have proposed the modification in these approaches by incorporating temporal dimension. The new approach was compared with their non-temporal counterparts and the results were analyzed for classifier accuracy and execution time. The study concludes that temporal CPAR achieves better performance than temporal CBA and temporal CMAR. The three temporal associative classifiers (TACs) were compared on ten different datasets for classifier accuracy and significant conclusion was drawn as temporal associative classifiers performed better than their non-temporal counterparts, while temporal CPAR being the best among the three TACs.

Fuzzy TOPSIS Method Applied for Ranking of Teacher in Higher Education

Higher education institutions including technical institutions are facing problems for providing quality education. A survey reported that there are crises of good and qualified teachers in higher education system. To detain quality teachers and to reject others an exposited opinion is required, but due to conflicting criteria on them it is very difficult to decide rank of quality Teachers, hence a suitable techniques for selecting and ranking of existing teachers is required. Multicriteria decision making (MCDM) is a technique which can be used in this scenario. Fuzzy technique for order preference by similarity to ideal solution (FTOPSIS) is a MCDM method in which various criteria can be fuzzified using fuzzy logic to deal the problem in precise manner. In this research work, FTOPSIS method is applied on the sample data collected from different higher education intuitions and weights are obtained with the help of another MCDM method called analytic hierarchy process (AHP) method. A small sample of 10 teachers and 5 criteria are considered to demonstrate FTOPSIS method to find out ranking among them, obtained results are verified with those collected from various experts and found to be satisfactory.

Study of Recommendation System for Web Portals

Recommendation system (RS) is one of the most advanced approach which is popular commercially and in Research community. Many of the Web portals are using Recommender system to increase their customers and providing them better recommendation for purchasing of products. It learns from the customer’s behavior of purchasing, rating and commenting, then deciding the score by help of Recommender system. In this paper, introducing about Recommendation system and its various types with their corresponding technologies that are currently used in E-commerce web portals. Later explaining some of the well known portals using Recommender system and comparison in techniques. Paper conclude with the applications of recommendation system and how they are increasing customer’s to E-commerce.

A group expert evaluation for teachers by integrating fuzzy AHP and TOPSIS models

Teachers are an essential pillar of any educational institute. The selection of eminence teacher is a very important process to sustain the quality education. Selection of teachers depends on performance for selection criteria to generate a final ranking. The performance criteria of teachers are evaluated by experts and it is always ambiguous. This ambiguity of human decision-making can be deal with the fuzzy decision support model. This work is an attempt to construct a fuzzy AHP and TOPSIS model to assess different teachers and select quality teachers for educational intuitions. The proposed method enables decision analysts to better understand the complete evaluation process and provide a more precise, valuable, and systematic decision support tool.

Routing Techniques for Reliable Wireless Sensor Networks

Wireless Sensor Network (WSN) is a collection of thousands of tiny sensor nodes having capability of wireless communication, limited computation and sensing. It is now used in many application including military, environmental, healthcare application, home automation and traffic control. In this paper we will compare Data centric routing protocols for wireless sensor network. It also discusses about simulation based study of routing protocols such as Flooding and Directed Diffusion.