Volume 7 Number 2 September 2017


Misbehavior Nodes Detection in VANET Using Watchdog Techniques
S. Raagavi, S. Sathish

Abstract- VANET is a subset of Mobile Ad- hoc Networks (MANET) in which communication nodes are mainly vehicles. VANETS enable wireless communication between vehicles and vehicle to infrastructure. Its main objective is to render safety, comfort and convenience on the road. VANET is different from ad-hoc networks due to its unique characteristics. VANET being an ad-hoc network are at risk of various misbehaviors like tampering of messages, eavesdropping, spamming, masquerading itís because of the lack of centralized administration. Security of VANET has been identified as one of the major challenges. In order to do the watchdog correctly and effectively, it must follow the security requirements such as integrity, confidentiality, privacy, non reputation and authentication to protect against attackers and malicious vehicular nodes. Vehicular ad-hoc network relies on cooperation between vehicles and implemented the main techniques for watchdog used to detect the misbehavior node on vehicular communication. A misbehaving node may use to watchdog techniques transmit false alerts, tamper messages, create congestion in the network drop, delay and duplicate packets. Thus detecting misbehaving nodes in VANET is very crucial and indispensable as it might have disastrous consequences.


Score Based Co-Clustering for Binary Data
R.Gowri, R. Rathipriya

Abstract- Most of the datasets like medical datasets, expression data, network data, sensor datasets are in binary format. This article focuses on mining the block of oneís or zeroís (constant co-cluster) in the binary data. It represents the likelihood characteristics among the local group of elements in the data. For this purpose a score based co-clustering approach is proposed in this article. Initially this approach is attempted on the four different synthetic datasets under noisy and noiseless environments. The experimental results are compared with existing co-clustering approaches like BiMax and xMotif algorithms. The results evidence that, the proposed approach is performing well in mining the constant co-clusters in both noisy and noiseless environments.


Automatic Detection And Prediction Of Skin Cancer Using K-Means Clustering
R.Sylviya, R.Venkatachalam

Abstract- Skin Cancer is common disease among the people throughout the world which is very dangerous and affective too. The type of biopsy method identify the skin cancer is much harmful. Skin cancer being detected at earlier stages can save more million peoples. According to my clarification there is a device of automatic detection of medical digital image which can be cured at early stage of skin cancer. The skin cancer divides into two categories of non-melanoma and melanoma. These categories are used to check the texture and colour characters of the skin which gives us an enrich result. The segmented process in the skin cancer describes the injury of skin malignance which is used in k-means clustering process. The proposed classification is calculated on four distinct forms of classification rate. In these classification rates perform to provides the best possibility of the predicting the skin cancer of our proposed system that are used in the TDV value (Total Dermoscopic Value). The GLCM algorithm is used to feature extraction of the skin cancer i.e. bio-digital image.


Query Based Tri-Clustering (QBTC)
N. Narmadha

Abstract- In the world of big data, accumulated 3D Gene Expression Data is increased rapidly, therefore a novel query based tricluster is proposed in this work to extract maximum similarity tricluster from the given 3D data. The main advantage of this proposed work is query tricluster is the identification of customized tricluster with respect to the given query. This query is the most valuable or functionable gene. The performance of the proposed work is studied with the stimulated data. It has observed the query Tri performance well in extracting constant, shifting and scaling pattern tricluster.


Attribute Selection using Machine Learning Technique

Abstract- A Central problem in machine learning is to identify a representative set of attributes from which to construct a classification model for a particular task. Attribute selection is a well-known problem in the field of machine learning technique. It allows probabilistic classification and shows promising results on several benchmark problems. Attribute Selection is a task of choosing a small subset of features/attributes that is sufficient to predict the target labels well. Attribute Selection reduces the computational complexity of learning and prediction algorithms and saves computational the cost spent for measuring irrelevant features. This work addresses the problem of attribute selection for machine learning through Regression Analysis with different attribute selection methods like Forward Selection, Backward Elimination and Quick Reduct algorithm. The performance of the proposed approaches is studied based on the AIC measure. Further the classification accuracy of the proposed approach is analyzed by comparing it with the benchmark classification algorithm like K-Nearest Neighbor approach and Decision Tree approach. The result shows that accuracy of the classification algorithm without attribute selection. The proposed approach greatly improves the efficiency of the classification algorithms and the prediction accuracy is also remains satisfactory. So the Quick Reduct based attribute selection is better for machine learning techniques.