Cristina Kadar

Automatic Classification of Change Requests for Improved IT Service Quality

Faulty changes to the IT infrastructure can lead to critical system and application outages, and therefore cause serious economical losses. In this paper, we describe a change planning support tool that aims at assisting the change requesters in leveraging aggregated information associated with the change, like past failure reasons or best implementation practices. The thus gained knowledge can be used in the subsequent planning and implementation steps of the change. Optimal matching of change requests with the aggregated information is achieved through the classification of the change request into about 200 fine-grained activities. We propose to automatically classify the incoming change requests using various information retrieval and machine learning techniques. The cost of building the classifiers is reduced by employing active learning techniques or by leveraging labeled features. Historical tickets from two customers were used to empirically assess and compare the accuracy of the different classification approaches (Lucene index, multinomial logistic regression, and generalized expectation criteria).

CityWatch: the personalized crime prevention assistant

Motivated by rising levels of crime against property and findings in criminology research, we are developing CityWatch - the first mobile application that supports crime prevention behavior at community level. CityWatch leverages data on past crime incidents, which are sourced both from trustworthy sources, like the national census and the insurance industry, and from its users through crowd-sourcing. It applies machine learning algorithms to analyze the past incidents together with further data characterizing the living areas and learns common patterns of crime. These patterns are then leveraged in a general forecasting component, as well as in generating personalized risk profiles and crime prevention tips for registered users based on their account information. The results are visualized in an interactive map, where users can analyze past crime in their neighborhood and view predictions of future crime. Users can report a new crime and opt to receive notifications about new incidents in their proximity or area of residence.

Measuring Ambient Population from Location-Based Social Networks to Describe Urban Crime

Recently, a lot of attention has been given to crime prediction, both by the general public and by the research community. Most of the latest work has concentrated on showing the potential of novel data sources like social media, mobile phone data, points of interest, or transportation data for the crime prediction task and researchers have focused mostly on techniques from supervised machine learning to show their predictive potential. Yet, the question remains if indeed this data can be used to better describe urban crime. In this paper, we investigate the potential of data harvested from location-based social networks (specifically Foursquare) to describe urban crime. Towards this end, we apply techniques from spatial econometrics. We show that this data, seen as a measurement for the ambient population of a neighborhood, is able to further describe crime levels in comparison to models built solely on census data, seen as measurement for the resident population of a neighborhood. In an analysis of crime on census tract level in New York City, the total number of incidents can be described by our models with up to \(R^2 = 56\%\), while the best model for the different crime subtypes is achieved for larcenies with roughly \(67\%\) of the variance explained.

Towards a crowdsourcing approach for crime prevention

With the rising level of criminal activities, crime is becoming one of the main problems of modern society. To address this issue, we implement a mobile application for crime prevention. We focus on the usage intention and motivations for content creation and consumption. Our results indicate that people are willing to use the app for acquiring and sharing crime-related information, but not on a daily basis. In addition, participation on the platform was found to be driven by affective and rational motivations, to contribute to the neighborhood safety and in return receive help for maintaining personal safety.

Domain adaptation for text categorization by feature labeling

We present a novel approach to domain adaptation for text categorization, which merely requires that the source domain data are weakly annotated in the form of labeled features. The main advantage of our approach resides in the fact that labeling words is less expensive than labeling documents. We propose two methods, the first of which seeks to minimize the divergence between the distributions of the source domain, which contains labeled features, and the target domain, which contains only unlabeled data. The second method augments the labeled features set in an unsupervised way, via the discovery of a shared latent concept space between source and target. We empirically show that our approach outperforms standard supervised and semi-supervised methods, and obtains results competitive to those reported by state-of-the-art domain adaptation methods, while requiring considerably less supervision.

Towards a Burglary Risk Profiler Using Demographic and Spatial Factors

According to modern crime victimization theories, the offender, the victim, and the spatial environment equally affect the likelihood of a crime getting committed, especially in the case of burglaries. With this in mind, we compile an extensive list of potential drivers of burglary by aggregating data from different open data sources, such as census statistics social, demographic, and economic data, points of interest, and the national road network. Based on the underlying data distribution, we build statistical models that automatically select the risk factors affecting the burglary numbers in the Swiss municipalities and predict the level of future crimes. The gained information is integrated in a crime prevention information system providing its users a view of the current crime exposure in their neighborhood.

Mining large-scale human mobility data for long-term crime prediction

Traditional crime prediction models based on census data are limited, as they fail to capture the complexity and dynamics of human activity. With the rise of ubiquitous computing, there is the opportunity to improve such models with data that make for better proxies of human presence in cities. In this paper, we leverage large human mobility data to craft an extensive set of features for crime prediction, as informed by theories in criminology and urban studies. We employ averaging and boosting ensemble techniques from machine learning, to investigate their power in predicting yearly counts for different types of crimes occurring in New York City at census tract level. Our study shows that spatial and spatio-temporal features derived from Foursquare venues and checkins, subway rides, and taxi rides, improve the baseline models relying on census and POI data. The proposed models achieve absolute R2

Towards Simulating Criminal Offender Movement Based on Insights from Human Dynamics and Location-Based Social Networks

R^{2}

Automatically Identifying Tag Types

metrics of up to 65% (on a geographical out-of-sample test set) and up to 89% (on a temporal out-of-sample test set). This proves that, next to the residential population of an area, the ambient population there is strongly predictive of the area’s crime levels. We deep-dive into the main crime categories, and find that the predictive gain of the human dynamics features varies across crime types: such features bring the biggest boost in case of grand larcenies, whereas assaults are already well predicted by the census features. Furthermore, we identify and discuss top predictive features for the main crime categories. These results offer valuable insights for those responsible for urban policy or law enforcement.

Digital Neighborhood Watch: To share or not to share?

Interest in data-driven crime simulations has been growing in recent years, confirming its potential to advance crime prevention and prediction. Especially, the use of new data sources in crime simulation models can contribute towards safer and smarter cities. Previous work on agent-based models for crime simulations have intended to simulate offender behavior in a geographical environment, relying exclusively on a small sample of offender homes and crime locations. The complex dynamics of crime and the lack of information on criminal offender’s movement patterns challenge the design of offender movement in simulations. At the same time, the availability of big, GPS-based user data samples (mobile data, social media data, etc.) already allowed researchers to determine the laws governing human mobility patterns, which, we argue, could inform offender movement. In this paper, we explore: (1) the use of location-based venue data from Foursquare in New York City (NYC), and (2) human dynamics insights from previous studies to simulate offender movement. We study 9 offender mobility designs in an agent-based model, combining search distances strategies (static, uniform distributed, and Levy-flight approximation) and target selection algorithms (random intersection, random Foursquare venues, and popular Foursquare venues). The offender behavior performance is measured using the ratio of crime locations passed vs average distance traveled by each offender. Our initial results show that agents moving between POI perform best, while the performance of the three search distance strategies is similar. This work provides a step forward towards more realistic crime simulations.