Biblio

In this paper, we present a multi-featured supervised automatic keyword extraction system. We extracted salient semantic features which are descriptive of candidate keyphrases, a Random Forest classifier was used for training. The system achieved an accuracy of 58.3 % precision and has shown to outperform two top performing systems when benchmarked on a crowdsourced dataset. Furthermore, our approach achieved a personal best Precision and F-measure score of 32.7 and 25.5 respectively on the Semeval Keyphrase extraction challenge dataset. The paper describes the approaches used as well as the result obtained.

Most anti-phishing solutions that exist today require scanning a large portion of the web, which is vast and equivalent to finding a needle in a haystack. In addition, such solutions are not very efficient. We propose a different approach. Our solution does not rely on the scanning of the entire Internet or a large portion of it and only needs access to the brand's traffic in order to be able to detect phishing attempts against that brand. By analyzing a sample of phishing websites, we find features that can be used to distinguish phishing websites from the legitimate ones. We then use these features to train a machine learning classifier capable of helping brands detect phishing attempts against them. Our approach can detect up to 86% of phishing attacks against the brands and is best used as a complementary tool to the existing anti-phishing solutions.

Record linkage refers to the task of finding same entity across different databases. We propose a machine learning based record linkage algorithm for financial entity databases. Record linkage on financial databases are essential for information integration on certain financial entity, since those databases do not have common unified identifier. Our algorithm works in two steps to determine if a pair of record is same entity or not. First we check with proposed rules if the record pair can be exactly matched after cleaning the entity name and address. Second, inspired by earlier work on author name disambiguation, we train a binary Random Forest classifier to decide the linkage. To reduce and scale the computation, this process is done only for candidate pairs within a proposed heuristic. Initial evaluation for precision, recall and F1 measures on two different linking tasks in the Financial Entity Identification and Information Integration (FEIII) Challenge show promising results.

Nowadays, a typical household owns multiple digital devices that can be connected to the Internet. Advertising companies always want to seamlessly reach consumers behind devices instead of the device itself. However, the identity of consumers becomes fragmented as they switch from one device to another. A naive attempt is to use deterministic features such as user name, telephone number and email address. However consumers might refrain from giving away their personal information because of privacy and security reasons. The challenge in ICDM2015 contest is to develop an accurate probabilistic model for predicting cross-device consumer identity without using the deterministic user information. In this paper we present an accurate and scalable cross-device solution using an ensemble of Gradient Boosting Decision Trees (GBDT) and Random Forest. Our final solution ranks 9th both on the public and private LB with F0.5 score of 0.855.