Satyen Abrol

Tweethood: Agglomerative Clustering on Fuzzy k-Closest Friends with Variable Depth for Location Mining

According to a recent report by research firm ABI Research, location-based social networks could reach revenues as high as

TWinner: understanding news queries with geo-content using Twitter

13.3 billion by 2014 [1]. Social Networks like Foursquare and Gowalla are in a dead heat in the Location War. But, having said that it is important to understand for privacy and security reasons, most of the people on social networking sites like Twitter are unwilling to specify their locations explicitly. This creates a need for software that mines the location of the user based on the implicit attributes associated with him. In this paper, we propose the development of a tool TweetHood that predicts the location of the user on the basis of his social network. We show the evolution of the algorithm, highlighting the drawbacks of the different approaches and our methodology to overcome them. We perform extensive experiments to show the validity of our system in terms of both accuracy and running time. The experiments performed demonstrate that our system achieves an accuracy of 72.1% at the city level and 80.1% at the country level. Experimental results show that TweetHood outperforms the gazetteer based geo-tagging approach.

Tweeque: Spatio-Temporal Analysis of Social Networks for Location Mining Using Graph Partitioning

In the present world scenario, where the search engines wars are becoming fiercer than ever, it becomes necessary for each search engine to realize the intent of the user query to be able to provide him with more relevant search results. Amongst the various categories of search queries, a major portion is constituted by those having news intent. Seeing the tremendous growth of social media users, the spatial-temporal nature of the media can prove to be a very useful tool to improve the search quality. In our work we examine the development of such a tool that combines social media in improving the quality of web search and predicting whether the user is looking for news or not. We go one step beyond the previous research by mining Twitter messages, assigning weights to them and determining keywords that can be added to the search query to act as pointers to the existing search engine algorithms suggesting to it that the user is looking for news. We conduct a series of experiments and show the impact that TWinner has on the results.

Design and implementation of SNODSOC: Novel class detection for social network analysis

Because of privacy and security reasons, most of the people on social networking sites like Twitter are unwilling to specify their locations in the profiles. In this paper, we present a completely novel approach, Tweeque which is a spatio-temporal mining algorithm that predicts the current location of the user purely on the basis of his social network. The algorithm goes beyond the previous approaches by linking geospatial proximity to friendship and understanding the social phenomenon of migration. The algorithm then performs graph partitioning for identifying social groups allowing us to implicitly consider time as a factor for prediction of users most current city location. We perform extensive experiments to show the validity of our system in terms of both accuracy and running time.

Real-Time Stream Data Analytics for Multi-purpose Social Media Applications

This paper describes a framework, SNODSOC (Stream based novel class detection for social network analysis), that detects evolving patterns and trends in social microblogs. SNODSOC extends our powerful data mining system, SNOD (Stream-based Novel Class Detection) for now detecting novel patterns and trends within microblogs.

MapIt: a case study for location driven knowledge discovery and mining

This paper describes a real-time information integration and analytics system called InXite for multi-purpose applications. InXite is designed to detect evolving patterns and trends in streaming data including social media data (e.g., tweets). InXite comprises of multiple modules including InXite Registration and Dashboard, InXite Real-time Data Streamer, InXite Information Integrator and InXite Analytics Engine each of which is outlined in this document. At present, InXite has been demonstrated for security, marketing and law enforcement applications.

Design and Implementation of a Data Mining System for Malware Detection

In the present world scenario, everybody is on the lookout for suitable housing options, each having different needs (e.g., the elderly are looking for safe, quiet neighbourhood, while students are looking for affordable apartments close to the university/school). For e.g., Craigslist currently does not have a map version, making the process of apartment searching a very long and laborious process. This creates a need for software that is significantly superior to current web search tools. We demonstrate the development of a tool which takes the Craigslist apartment listings on Google Maps. MapIt then integrates this functionality with the information collected from location based extraction of various web sources such as the city police blotter which makes apartment searching simpler and faster, helping the user to make a better decision. The paper also discusses the challenges that are faced in the development process, the raw and unstructured nature of the documents, the existence of geo/non-geo and geo-geo disambiguities and our approach in identifying the location of the apartment from informal text (geo-parsing and geo-tagging of content) to ensure maximum coverage of the listings.

Geospatial Schema Matching with High-Quality Cluster Assurance and Location Mining from Social Network

This paper describes the design and implementation of a data mining system called SNODMAL Stream based novel class detection for malware for malware detection. SNODMAL extends our data mining system called SNOD Stream-based Novel Class Detection for detecting malware. SNOD is a powerful system as it can detect novel classes. We also describe the design of SNODMAL++ which is an extended version of SNODMAL.

Tweecalization: Efficient and intelligent location mining in twitter using semi-supervised learning

In this talk, we will present how semantics can improve the quality of the data mining process. In particular, first, we will focus on geospatial schema matching with high quality cluster assurance. Next, we will focus on location mining from social network. With regard to the first problem, resolving semantic heterogeneity across distinct data sources remains a highly relevant problem in the GIS domain requiring innovative solutions. Our approach, called GSim, semantically aligns tables from respective GIS databases by first choosing attributes for comparison. We then examine their instances and calculate a similarity value between them called Entropy-Based Distribution (EBD) by combining two separate methods. Our primary method discerns the geographic types from instances of compared attributes. If geographic type matching is not possible, we then apply a generic schema matching method which employs normalized Google distance with the usage of clustering process. GSim proceeds by deriving clusters from attribute instances based on content and their geographic types (if possible), gleaned from a gazetteer. However, clustering algorithms may produce inconsistent results based on variable cluster quality. We apply novel metrics measuring cluster distance and purity to guarantee high-quality homogeneous clusters. The end result is a wholly geospatial similarity value, expressed as EBD. We show the effectiveness of our approach over the traditional N-gram approach across multi-jurisdictional datasets by generating impressive results. With regard to the second problem, we will predict the location of the user on the basis of his social network (e.g., Twitter) using the strong theoretical framework of semi-supervised learning, in particular, we employ label propagation algorithm. For privacy and security reasons, most of the people on social networking sites like Twitter are unwilling to specify their locations explicitly. On the city locations returned by the algorithm, the system performs agglomerative clustering based on geospatial proximity and their individual scores to return cluster of locations with higher confidence. We perform extensive experiments to show the validity of our system in terms of both accuracy and running time. Experimental results show that our approach outperforms the content based geo-tagging approach in both accuracy and running time.