A typical problem for a search engine (hosting sponsored search service) is to provide the advertisers with a forecast of the number of impressions his/her ad is likely to obtain for a given bid. Accurate forecasts have high business value, since they enable advertisers to select bids that lead to better returns on their investment. They also play an important role in services such as automatic campaign optimization. Despite its importance the problem has remained relatively unexplored in literature. Existing methods typically overfit to the training data, leading to inconsistent performance. Furthermore, some of the existing methods cannot provide predictions for new ads, i.e., for ads that are not present in the logs. In this paper, we develop a generative model based approach that addresses these drawbacks. We design a Bayes net to capture inter-dependencies between the query traffic features and the competitors in an auction. Furthermore, we account for variability in the volume of query traffic by using a dynamic linear model. Finally, we implement our approach on a production grade MapReduce framework and conduct extensive large scale experiments on substantial volumes of sponsored search data from Bing. Our experimental results demonstrate significant advantages over existing methods as measured using several accuracy/error criteria, improved ability to provide estimates for new ads and more consistent performance with smaller variance in accuracies. Our method can also be adapted to several other related forecasting problems such as predicting average position of ads or the number of clicks under budget constraints.

We present alignment refinement by mining coupled residues (ARMiCoRe), a novel approach to a classical bioinformatics problem, viz., multiple sequence alignment (MSA) of gene and protein sequences. Aligning multiple biological sequences is a key step in elucidating evolutionary relationships, annotating newly sequenced segments, and understanding the relationship between biological sequences and functions. Classical MSA algorithms are designed to primarily capture conservations in sequences whereas couplings, or correlated mutations, are well known as an additional important aspect of sequence evolution. (Two sequence positions are coupled when mutations in one are accompanied by compensatory mutations in another). As a result, better exposition of couplings is sometimes one of the reasons for hand-tweaking of MSAs by practitioners. ARMiCoRe introduces a distinctly pattern mining approach to improving MSAs: using frequent episode mining as a foundational basis, we define the notion of a coupled pattern and demonstrate how the discovery and tiling of coupled patterns using a max-flow approach can yield MSAs that are better than conservation-based alignments. Although we were motivated to improve MSAs for the sake of better exposing couplings, we demonstrate that our MSAs are also improvements in terms of traditional metrics of assessment. We demonstrate the effectiveness of ARMiCoRe on a large collection of data sets.

Discovering frequent episodes over event sequences is an important data mining problem. Existing methods typically require multiple passes over the data, rendering them unsuitable for streaming contexts. We present the first streaming algorithm for mining frequent episodes over a window of recent events in the stream. We derive approximation guarantees for our algorithm in terms of: (i) the separation of frequent episodes from infrequent ones, and (ii) the rate of change of stream characteristics. Our parameterization of the problem provides a new sweet spot in the tradeoff between making distributional assumptions over the stream and algorithmic efficiencies of mining. We illustrate how this yields significant benefits when mining practical streams from neuroscience and telecommunications logs.

This paper presents the first evaluation framework for Web search query segmentation based directly on IR performance. In the past, segmentation strategies were mainly validated against manual annotations. Our work shows that the goodness of a segmentation algorithm as judged through evaluation against a handful of human annotated segmentations hardly reflects its effectiveness in an IR-based setup. In fact, state-of the-art algorithms are shown to perform as good as, and sometimes even better than human annotations a fact masked by previous validations. The proposed framework also provides us an objective understanding of the gap between the present best and the best possible segmentation algorithm. We draw these conclusions based on an extensive evaluation of six segmentation strategies, including three most recent algorithms, vis-a-vis segmentations from three human annotators. The evaluation framework also gives insights about which segments should be necessarily detected by an algorithm for achieving the best retrieval results. The meticulously constructed dataset used in our experiments has been made public for use by the research community.

Discovering frequent patterns from data is a popular exploratory technique in datamining. However, if the data are sensitive (e.g., patient health records, user behavior records) releasing information about significant patterns or trends carries significant risk to privacy. This paper shows how one can accurately discover and release the most significant patterns along with their frequencies in a data set containing sensitive information, while providing rigorous guarantees of privacy for the individuals whose information is stored there.

Mining temporal network models from discrete event streams is an important problem with applications in computational neuroscience, physical plant diagnostics, and human-computer interaction modeling. We focus in this paper on temporal models representable as excitatory networks where all connections are stimulative, rather than inhibitory. Through this emphasis on excitatory networks, we show how they can be learned by creating bridges to frequent episode mining. Specifically, we show that frequent episodes help identify nodes with high mutual information relationships and which can be summarized into a dynamic Bayesian network (DBN). To demonstrate the practical feasibility of our approach, we show how excitatory networks can be inferred from both mathematical models of spiking neurons as well as real neuroscience datasets.

This paper presents a new algorithm for sequence prediction over long categorical event streams. The input to the algorithm is a set of target event types whose occurrences we wish to predict. The algorithm examines windows of events that precede occurrences of the target event types in historical data. The set of significant frequent episodes associated with each target event type is obtained based on formal connections between frequent episodes and Hidden Markov Models (HMMs). Each significant episode is associated with a specialized HMM, and a mixture of such HMMs is estimated for every target event type. The likelihoods of the current window of events, under these mixture models, are used to predict future occurrences of target events in the data. The only user-defined model parameter in the algorithm is the length of the windows of events used during model estimation. We first evaluate the algorithm on synthetic data that was generated by embedding (in varying levels of noise) patterns which are preselected to characterize occurrences of target events. We then present an application of the algorithm for predicting targeted user-behaviors from large volumes of anonymous search session interaction logs from a commercially-deployed web browser tool-bar.

Frequent itemsets mining is a popular framework for pattern discovery. In this framework, given a database of customer transactions, the task is to unearth all patterns in the form of sets of items appearing in a sizable number of transactions. We present a class of models called Itemset Generating Models (or IGMs) that can be used to formally connect the process of frequent item- sets discovery with the learning of generative models. IGMs are specified using simple probability mass functions (over the space of transactions), peaked at specific sets of items and uniform everywhere else. Under such a connection, it is possible to rigorously associate higher frequency patterns with generative models that have greater data likelihoods. This enables a generative model-learning interpretation of frequent itemsets mining. More importantly, it facilitates a statistical significance test which prescribes the minimum frequency needed for a pattern to be considered interesting. We illustrate the effectiveness of our analysis through experiments on standard benchmark data sets.

Frequent episode discovery is a popular framework for mining data available as a long sequence of events. An episode is essentially a short ordered sequence of event types and the frequency of an episode is some suitable measure of how often the episode occurs in the data sequence. Recently,we proposed a new frequency measure for episodes based on the notion of non-overlapped occurrences of episodes in the event sequence, and showed that, such a definition, in addition to yielding computationally efficient algorithms, has some important theoretical properties in connecting frequent episode discovery with HMM learning. This paper presents some new algorithms for frequent episode discovery under this non-overlapped occurrences-based frequency definition. The algorithms presented here are better (by a factor of N, where N denotes the size of episodes being discovered) in terms of both time and space complexities when compared to existing methods for frequent episode discovery. We show through some simulation experiments, that our algorithms are very efficient. The new algorithms presented here have arguably the least possible orders of spaceand time complexities for the task of frequent episode discovery.

This paper is concerned with the framework of frequent episode discovery in event sequences. A new temporal pattern, called the generalized episode, is defined, which extends this framework by incorporating event duration constraints explicitly into the pattern's definition. This new formalism facilitates extension of the technique of episodes discovery to applications where data appears as a sequence of events that persist for different durations (rather than being instantaneous). We present efficient algorithms for episode discovery in this new framework. Through extensive simulations, we show the expressive power of the new formalism. We also show how the duration constraint possibilities can be used as a design choice to properly focus the episode discovery process. Finally, we briefly discuss some interesting results obtained on data from manufacturing plants of General Motors.

This paper establishes a formal connection between two common, but previously unconnected methods for analyzing data streams: discovering frequent episodes in a computer science framework and learning generative models in a statistics framework. We introduce a special class of discrete hidden Markov models (HMMs), called episode generating HMMs (EGHs), and associate each episode with a unique EGH. We prove that, given any two episodes, the EGH that is more likely to generate a given data sequence is the one associated with the more frequent episode. To be able to establish such a relationship, we define a new measure of frequency of an episode, based on what we call nonoverlapping occurrences of the episode in the data. An efficient algorithm is proposed for counting the frequencies for a set of episodes. Through extensive simulations, we show that our algorithm is both effective and more efficient than current methods for frequent episode discovery. We also show how the association between frequent episodes and EGHs can be exploited to assess the significance of frequent episodes discovered and illustrate empirically how this idea may be used to improve the efficiency of the frequent episode discovery.