P. Radha Krishna

A cryptography based privacy preserving solution to mine cloud data

Due to increased adoption of cloud computing, there is a growing need of addressing the data privacy during mining. On the other hand, knowledge sharing is a key to survive many business organizations. Several attempts have been made to mine the data in distributed environment however, maintaining the privacy while mining the data over cloud is a challenging task. In this paper, we present an efficient and practical cryptographic based scheme that preserves privacy and mine the cloud data which is distributed in nature. In order to address the classification task, our approach uses k-NN classifier. We extend the Jaccard measure to find the similarity between two encrypted and distributed records by conducting an equality test. In addition, our approach accelerates mining by finding nearest neighbours at local and then at global level. The proposed approach avoids transmitting the original data and sharing of the key that is required in traditional crypto based privacy preserving data mining solutions.

Preemption-Aware Energy Management in Virtualized Data Centers

Energy efficiency is one of the main challenge hat data centers are facing nowadays. A considerable portion of the consumed energy in these environments is wasted because of idling resources. To avoid wastage, offering services with variety of SLAs (with different prices and priorities) is a common practice. The question we investigate in this research is how the energy consumption of a data center that offers various SLAs can be reduced. To answer this question we propose an adaptive energy management policy that employs virtual machine(VM) preemption to adjust the energy consumption based on user performance requirements. We have implementedour proposed energy management policy in Haize a as a real scheduling platform for virtualized data centers. Experimental results reveal 18% energy conservation (up to 4000 kWh in 30 days) comparing with other baseline policies without any major increase in SLA violation.

SeqPAM: A Sequence Clustering Algorithm for Web Personalization

With the growth in the number of Web users and necessity for making information available on the Web, the problem of Web personalization has become very critical and popular. Developers are trying to customize a Web site to the needs of specific users with the help of knowledge acquired from user navigational behavior. Since user page visits are intrinsically sequential in nature, efficient clustering algorithms for sequential data are needed. In this chapter, we introduce a similarity preserving function called sequence and set similarity measure S3M that captures both the order of occurrence of page visits as well as the content of pages. We conducted pilot experiments comparing the results of PAM, a standard clustering algorithm, with two similarity measures: Cosine and S3M. The goodness of the clusters resulting from both the measures was computed using a cluster validation technique based on average levensthein distance. Results on pilot dataset established the effectiveness of S3M for sequential data. Based on these results, we proposed a new clustering algorithm, SeqPAM for clustering sequential data. We tested the new algorithm on two datasets namely, cti and msnbc datasets. We provided recommendations for Web personalization based on the clusters obtained from SeqPAM for msnbc dataset.

Pattern Discovery Using Sequence Data Mining: Applications and Studies

Sequential data from Web server logs, online transaction logs, and performance measurements is collected each day. This sequential data is a valuable source of information, as it allows individuals to search for a particular value or event and also facilitates analysis of the frequency of certain events or sets of related events. Finding patterns in sequences is of utmost importance in many areas of science, engineering, and business scenarios.Pattern Discovery Using Sequence Data Mining: Applications and Studies provides a comprehensive view of sequence mining techniques and presents current research and case studies in pattern discovery in sequential data by researchers and practitioners. This research identifies industry applications introduced by various sequence mining approaches.

A privacy preserving Jaccard similarity function for mining encrypted data

Due to advances in data collection and increasing dependency on data mining experts, preserving privacy of the data is a major concern when mining the data. Most of the classifier implementations for data mining have the tradeoff between classification accuracy and maintenance of data privacy. Another important aspect in distance-based classifiers is to accurately compute distance (or similarity) between two or more data points. In privacy preserving data mining techniques, providing a suitable distance measure to classify the data while maintaining data privacy is a challenging task. In this paper, we present an approach to compute similarity between two encrypted data points. We augmented Jaccard similarity function with Private Equality Test protocol facilitating a semi honest third party to conduct the equality test. The proposed privacy preserving scheme provides an efficient mechanism for similarity computation with reduced communication cost for mining the data.

Mining e-contract documents to classify clauses

E-contracts begin as legal documents and end up as processes that help organizations abide by legal rules while fulfilling contract terms. As contracts are complex, their deployment is predominantly established and fulfilled with significant human involvement. One of the key difficulties with any kind of contract processing is the legal ambiguity, which makes it difficult to address any violation of the contract terms. Thus, there is a need to track clauses for the contract activities under execution and violation of clauses. This necessitates deriving clause patterns from e-contract documents and map to their respective activities for further monitoring and fulfillment of e-contracts during their enactment. In this paper, we present a classification approach to extract clause patterns from e-contract documents. This is a challenging task as activities and clauses are mostly derived from both legal and business process driven contract knowledge.

Analyzing Internet Slang for Sentiment Mining

Every consumer has his own opinion about the product he is using which they are willing to share in social groups like forums, chat rooms and weblogs. As these review comments are actual feedbacks from customers, mining the sentiments in these reviews is being increasingly inducted into the feedback pipeline for any company. Along with it, the increasing use of slang in such communities in expressing emotions and sentiment makes it important to consider Slang in determining the sentiment. In this paper, we present an approach for finding the sentiment score of newly found slang sentiment words in blogs, reviews and forum texts over the World Wide Web. A simple mechanism for calculating sentiment score of documents using slang words with the help of Delta Term Frequency and Weighted Inverse Document Frequency technique is also presented in this paper.

On-road vehicle detection by cascaded classifiers

We present an efficient algorithm for on-road vehicle (e.g. side and rear view of cars) detection problem using cascade of boosted classifiers. Adaptive boosting based classifier in cascaded structure is one of the few good approaches for object detection. This approach filters different non-target (negative) samples in different stages of cascaded structure according to their level of similarity with target object class. The boosted weak learners are quick and efficient for initial stages only, but in later stage of cascaded structure they are not efficient enough to remove the critical false alarms. In this paper, we propose a method of cascading complex features at the later stage of cascaded classifier to enhance the detection performance. We compared the performance of local and global texture features in combination with boosted haar like features. The best performance for on-road obstacle detection is achieved by Adaboost with Haar-like feature along with SVM and Histograms of Oriented Gradients (HOG) features.

Natural language querying over databases using cascaded CRFs

Retrieving information from relational databases using a natural language query is a challenging task. Usually, the natural language query is transformed into its approximate SQL or formal languages. However, this requires knowledge about database structures, semantic relationships, natural language constructs and also handling ambiguities due to overlapping column names and column values. We present a machine learning based natural language search system to query databases without any knowledge of Structure Query Language (SQL) for underlying database. The proposed system - Cascaded Conditional Random Field is an extension to Conditional Random Fields, an undirected graph model. Unlike traditional Conditional Random Field models, we offer efficient labelling schemes to realize enhanced quality of search results. The system uses text indexing techniques as well as database constraint relationships to identify hidden semantic relationships present in the data. The presented system is implemented and evaluated on two real-life datasets.

A community driven social recommendation system

Recommendation systems play an important role in suggesting relevant information to users. In this paper, we introduce community-wise social interactions as a new dimension for recommendations and present a social recommendation system using collaborative filtering and community detection approaches. We use (i) community detection algorithm to extract friendship relations among users by analyzing user-user social graph and (ii) user-item based collaborative filtering for rating prediction. We developed our approach using map-reduce framework. Our approach improves scalability, coverage and cold start issue of collaborative filtering based recommendation system. We carried out experiments on MovieLens and Facebook datasets, to predict the rating of the movie and produce top-k recommendations for new (cold start) user. The results are compared with traditional collaborative filtering based recommendation system.