As people increasingly rely on road maps in the digital age, manually maintained maps cannot keep up with the demand for accuracy and freshness, evidenced by the recent iOS map incident and the bidding war for Waze. There are many research works on automatic map inference using GPS data, and some have suggested that Google and Waze automate their map update processes to some degree with user data. However, existing published work focuses on refining road geometry. In reality, connectivity issues at intersections, including missing connections and unmarked turn restrictions, are much more prevalent and also more difficult to infer. In this paper, we report our study on the connectivity issues in the OSM Shanghai map using 21 months of GPS data from over |$10\,000$| taxis. We first adapt a robust map matching algorithm. Then, we propose simple algorithms for detecting missing intersections, turn restrictions and road closures. Empirical results show that our algorithms of connectivity refinement for road maps are effective.

As people increasingly rely on road maps in the digital age, manually maintained maps cannot keep up with the demand for accuracy and freshness, evidenced by the recent iOS map incident and the bidding war for Waze. There are many research works on automatic map inference using GPS data, and some have suggested that Google and Waze automate their map update processes to some degree with user data. However, existing published work focuses on refining road geometry. In reality, connectivity issues at intersections, including missing connections and unmarked turn restrictions, are much more prevalent and also more difficult to infer. In this paper, we report on our study on the connectivity issues in the OSM Shanghai map using 21 months of GPS data from over 10, 000 taxis. We first adapt a robust map matching algorithm to detect missing intersections, and then train a time-series detection model for every turn possibility of every intersection using supervised learning.

The process of map matching takes a sequence of possibly noisy GPS coordinates from a vehicle trace and estimates the actual road positions---a crucial first step needed by many GPS applications. There has been a plethora of methods for map matching published, but most of them are evaluated on low-noise datasets obtained from a planned route. And comparisons with other methods are very limited. Based on our previous unifying framework used to catalog different mathematical formulas in many published methods, we evaluate representative algorithms using the low-noise dataset from the GIS Cup 2012 and a high-noise dataset collected from Shanghai downtown. Our experiments reveal that global max-weight and global geometrical map matching methods are the most accurate, but each has its weaknesses. We therefore propose a new map matching algorithm that integrates Fréchet distance with global weight optimization, which is more accurate across all sampling intervals.

The inaccuracy of manually created digital road maps is a persistent problem, despite their high economic value. We present CrowdAtlas, which automates map update based on people's travels, either individually or crowdsourced. Its mobile navigation app detects significant portions of GPS traces that do not conform to the existing map, as determined by state-of-the-art Viterbi map matching. When there is sufficient evidence collected, map inference algorithms can automatically update the map. The CrowdAtlas server aggregates exceptional traces from users with the navigation app as well as from other, large-scale data sources. From these it automatically generates high quality map updates, which can be propagated to its navigation app and other interested applications. Using CrowdAtlas app, we mapped out a 4.5 km^2 street block in Shanghai in less than half an hour and built a walking/cycling map of the SJTU campus. Using taxi traces collected from Beijing, we contributed completely computer-generated roads for this large, 61 km of missing roads to OpenStreetMap, the first set of open-source map community.

Digital road maps have become essential to many aspects of our lives. Unfortunately, they have persistent quality issues, both in developing countries as well as in developed countries, evidenced by the recent Apple-Google map war. A survey of British drivers showed 26% have been directed by their GPS to go into no-entry areas, and the news periodically reports car accidents caused by or related to digital maps. In addition to correcting existing errors, maps need to be frequently updated to reflect the latest constructions, closures, and reconfigurations. TomTom estimates that roads change by as much as 15% each year. There are also growing demands for other types of maps, including off-road driving, cycling, hiking, and skiing maps. There are services for people to share GPS traces, but none creates navigable maps.

This paper describes a map matching program submitted to the ACM SIGSPATIAL Cup 2012. We first summarize existing map matching algorithms into three categories, and compare their performance thoroughly. In general, global max-weight methods using the Viterbi dynamic programming algorithm are the most accurate but the accuracy varies at different sampling intervals using different weight functions. Our submission selects a hybrid that improves upon the best two weight functions such that its accuracy is better than both and the performance is robust against varying sampling rates. In addition, we employ many optimization techniques to reduce the overall latency, as the scoring heavily emphasizes on speed. Using the training dataset with manually corrected ground truth, our Java-based program matched all 14,436 samples in 5 seconds on a dual-core 3.3 GHz iCore 3 processor, and achieved 98.9% accuracy.

We address the problem of inferring road maps from large-scale GPS traces that have relatively low resolution and sampling frequency. Unlike past published work that requires high-resolution traces with dense sampling, we focus on situations with coarse granularity data, such as that obtained from thousands of taxis in Shanghai, which transmit their location as seldom as once per minute. Such data sources can be made available inexpensively as byproducts of existing processes, rather than having to drive every road with high-quality GPS instrumentation just for map building - and having to re-drive roads for periodic updates. Although the challenges in using opportunistic probe data are significant, successful mining algorithms could potentially enable the creation of continuously updated maps at very low cost.

Feature filtering methods are often used in text classification and other high-dimensional domains to quickly score each feature independently and pass only the best to the learning algorithm. The panoply of available methods grows over the years, with frequent research publications touting new functions that seem to yield superior learning: variants on Information Gain, Chi Squared, Mutual Information, and others. This slow generate-and-test search in the literature is usually counted as progress towards finding superior filter methods. But this is illusory.

Using a clickstream sample of 2 billion URLs from many thousand volunteer Web users, we wish to analyze typical usage of keyword searches across the Web. In order to do this, we need to be able to determine whether a given URL represents a keyword search and, if so, which field contains the query. Although it is easy to recognize 'q' as the query field in 'http://www.google.com/search?hl=en&q=music', we must do this automatically for the long tail of diverse websites. This problem is the focus of this paper. Since the names, types and number of fields differ across sites, this does not conform to traditional text classification or to multi-class problem formulations. The problem also exhibits highly non-uniform importance across websites, since traffic follows a Zipf distribution.

Linear classifiers have been shown to be effective for many discrimination tasks. Irrespective of the learning algorithm itself, the final classifier has a weight to multiply by each feature. This suggests that ideally each input feature should be linearly correlated with the target variable (or anti-correlated), whereas raw features may be highly non-linear. In this paper, we attempt to re-shape each input feature so that it is appropriate to use with a linear weight and to scale the different features in proportion to their predictive value. We demonstrate that this pre-processing is beneficial for linear SVM classifiers on a large benchmark of text classification tasks as well as UCI datasets.

In the realm of machine learning for text classification, TF-IDF is the most widely used representation for real-valued feature vectors. However, IDF is oblivious to the training class labels and naturally scales some features inappropriately. We replace IDF with Bi-Normal Separation (BNS), which has been previously found to be excellent at ranking words for feature selection filtering. Empirical evaluation on a benchmark of 237 binary text classification tasks shows substantially better accuracy and F-measure for a Support Vector Machine (SVM) by using BNS scaling. A wide variety of other feature representations were later tested and found inferior, as well as binary features with no scaling. Moreover, BNS scaling yielded better performance without feature selection, obviating the need for feature selection.

Most research in speeding up text mining involves algorithmic improvements to induction algorithms, and yet for many large scale applications, such as classifying or indexing large document repositories, the time spent extracting word features from texts can itself greatly exceed the initial training time. This paper describes a fast method for text feature extraction that folds together Unicode conversion, forced lowercasing, word boundary detection, and string hash computation. We show empirically that our integer hash features result in classifiers with equivalent statistical performance to those built using string word features, but require far less computation and less memory.

We combine the speed and scalability of information retrieval with the generally superior classification accuracy offered by machine learning, yielding a two-phase text classifier that can scale to very large document corpora. We investigate the effect of different methods of formulating the query from the training set, as well as varying the query size. In empirical tests on the Reuters RCV1 corpus of 806,000 documents, we find runtime was easily reduced by a factor of 27x, with a somewhat surprising gain in F-measure compared with traditional text classification.

This paper promotes a new task for supervised machine learning research: quantification—the pursuit of learning methods for accurately estimating the class distribution of a test set, with no concern for predictions on individual cases. A variant for cost quantification addresses the need to total up costs according to categories predicted by imperfect classifiers. These tasks cover a large and important family of applications that measure trends over time. The paper establishes a research methodology, and uses it to evaluate several proposed methods that involve selecting the classification threshold in a way that would spoil the accuracy of individual classifications. In empirical tests, Median Sweep methods show outstanding ability to estimate the class distribution, despite wide disparity in testing and training conditions. The paper addresses shifting class priors and costs, but not concept drift in general.

We discuss our experiences in analyzing customer-support issues from the unstructured free-text fields of technical-support call logs. The identification of frequent issues and their accurate quantification is essential in order to track aggregate costs broken down by issue type, to appropriately target engineering resources, and to provide the best diagnosis, support and documentation for most common issues. We present a new set of techniques for doing this efficiently on an industrial scale, without requiring manual coding of calls in the call center. Our approach involves (1) a new text clustering method to identify common and emerging issues; (2) a method to rapidly train large numbers of categorizers in a practical, interactive manner; and (3) a method to accurately quantify categories, even in the face of inaccurate classifications and training sets that necessarily cannot match the class distribution of each new month's data. We present our methodology and a tool we developed and deployed that uses these methods for tracking ongoing support issues and discovering emerging issues at HP.

Machine learning is the mainstay for text classification. However, even the most successful techniques are defeated by many real-world applications that have a strong time-varying component. To advance research on this challenging but important problem, we promote a natural, experimental framework-the Daily Classification Task-which can be applied to large time-based datasets, such as Reuters RCV1.In this paper we dissect concept drift into three main subtypes. We demonstrate via a novel visualization that the recurrent themes subtype is present in RCV1. This understanding led us to develop a new learning model that transfers induced knowledge through time to benefit future classifier learning tasks. The method avoids two main problems with existing work in inductive transfer: scalability and the risk of negative transfer. In empirical tests, it consistently showed more than 10 points F-measure improvement for each of four Reuters categories tested.

Hewlett-Packard has many millions of technical support documents in a variety of collections. As part of content management, such collections are periodically merged and groomed. In the process, it becomes important to identify and weed out support documents that are largely duplicates of newer versions. Doing so improves the quality of the collection, eliminates chaff from search results, and improves customer satisfaction.The technical challenge is that through workflow and human processes, the knowledge of which documents are related is often lost. We required a method that could identify similar documents based on their content alone, without relying on metadata, which may be corrupt or missing.We present an approach for finding similar files that scales up to large document repositories. It is based on chunking the byte stream to find unique signatures that may be shared in multiple files. An analysis of the file-chunk graph yields clusters of related files. An optional bipartite graph partitioning algorithm can be applied to greatly increase scalability.

Information Gain is a well-known and empirically proven method for high-dimensional feature selection. We found that it and other existing methods failed to produce good results on an industrial text classification problem. On investigating the root cause, we find that a large class of feature scoring methods suffers a pitfall: they can be blinded by a surplus of strongly predictive features for some classes, while largely ignoring features needed to discriminate difficult classes. In this paper we demonstrate this pitfall hurts performance even for a relatively uniform text classification task. Based on this understanding, we present solutions inspired by round-robin scheduling that avoid this pitfall, without resorting to costly wrapper methods. Empirical evaluation on 19 datasets shows substantial improvements.