Ticiana L. Coelho da Silva

A Framework for Online Mobility Pattern Discovery from Trajectory Data Streams

Trajectory pattern mining allows characterizing movement behavior, which leverages new applications and services. Most existing approaches analyse the whole object trajectory rather that the current movement. Besides existing approaches for online pattern discovery are restricted to instantaneous positions. Subsequently, they fail to capture the movement behaviour along time. By continuously tracking moving objects sub-trajectories at each time window, rather than just the last position, it becomes feasible to gain insight on the current behaviour, and potentially detect mobility patterns in real time. This demonstration presents a novel framework for online mobility pattern discovery in sub-trajectory data streams. Key innovations include: (i) Online discovery of mobility patterns and pattern evolution by tracking the sub-trajectories of moving objects, (ii) A novel structure, called micro-group, to represent the relationship among moving objects, and (iii) An incremental algorithm to maintain micro-groups and to capture their evolution on highly dynamic sub-trajectory data. We present various demonstration scenarios using a real data set.

Discovering frequent mobility patterns on moving object data

We consider the problem of efficiently discovering and detecting frequent mobility patterns on moving object data. Our proposed approach is key for mobility applications, such as applications that need to discover and explain movement patterns of a set of moving objects (e.g. traffic management, birds migration, disease spreading). In this sense, we developed a method that performs density based clustering on trajectory data at regular time intervals, then we analyze clusters evolution, which is characterized by appear, disappear, expand, shrink, split, merge and survive. To solve our problem, a graph-based representation called Graph Evolution Cluster over Time (Δevol) is described and an algorithm to generate the graph is also presented. Finally, we map our problem to the problem of discovering frequent graph paths on Δevol. Therefore, the frequent graph paths are the frequent sequence of evolution patterns that occurs in the dataset. We discuss a preliminary solution to this problem and present some experimental results. The results suggest that evolution patterns and their frequency can be effectively obtained through the proposed Δevol obtained from moving object data.

Behavioral Analysis for Child Protection in Social Network through Data Mining and Multiagent Systems

The Internet connects millions of people worldwide, enabling diverse ways of interaction and social organization. Online Social Networks such as Facebook, MySpace, and Twitter, have created a new form of socialization that can provide good experiences for users. However, such systems, as well as connecting people, expose their users lives to others, making them subject of exploitation in many ways. This work explores specifically children and teenagers degree of exposition on Facebook. Due to the risk offered in distinct layers of the Internet, the aim of this work is to develop a smart tool that helps to avoid the action of individuals that are possibly a risky for children and teenagers, users of the social network Facebook, applying Data Mining techniques in a Multiagent System.

G2P: A Partitioning Approach for Processing DBSCAN with MapReduce

One of the most important aspects to consider when computing large data sets is to distribute and parallelize the analysis algorithms. A distributed system presents a good performance if the workload is properly balanced. It is expected that the computing time is directly related to the processing time on the node where the processing takes longer. This paper aims at proposing a data partitioning strategy that takes into account partition balance and that is generic for spatial data. Our proposed solution is based on a grid model data structure that is further transformed into a graph partitioning problem, where we finally compute the partitions. Our proposed approach is used on the distributed DBSCAN algorithm and it is focused on finding density areas in a large data set using MapReduce. We call our approach G2P (Grid and Graph Partitioning) and we show via massive experiments that G2P presents great quality data partitioning for the distributed DBSCAN algorithm compared to the competitors. We believe that G2P is not only suitable for DBSCAN algorithm, but also to execute spatial join operations and distance based range queries to name to a few.

A message classifier based on multinomial Naive Bayes for online social contexts

Children and teenagers today are increasingly connected to the internet. The use by minors of social networks applications, and games that are connected to the internet offer the possibility of communication, can make them exposed to various threats. One of the most troubling threats is sexual abuse. Thus the objective of this project is to create a model for classifying messages, as normal or dangerous, according to the risk they present to the minor. In addition to integrating the developed model with a project that analyzes the behavior of minors in a social network (Facebook), and calculates the risk of the minor be a victim of sexual abuse. Finally, we use the model in the classification of messages obtained from a server of the game Minecraft, quite popular among children.

Real-time discovery of hot routes on trajectory data streams using interactive visualization based on GPU

Abstract With the increasing availability of location acquisition technologies, massive movement data are collected continuously in a streaming manner. These data are a valuable source to help transit agencies to monitor the routes with heavy traffic (hot routes) and to identify abnormal events that require immediate attention to better direct traffic. In this regard, visual analytics can help by combining automated analysis with interactive visualization for effective understanding, reasoning, and decision-making. Traditional approaches aggregate movement by employing the concept of time-window discretization and exploring an entire dataset. However, they can present inconsistencies in time and space with the real traffic dynamics. In this paper, we present a novel approach to discover hot routes in real time. Different from other existing approaches, our method tracks the evolution of the objects’ movement in real time. We believe that no other approach captures and keeps track of how the hot routes evolve in an incremental manner. Moreover, we conducted extensive experiments on real-world and simulated datasets to evaluate the effectiveness and performance of our method. The results demonstrate that our method scales linearly with the size of the dataset, and is able to deal with large datasets and with streams of high-sampling rates.

Estimating Reference Evapotranspiration using Data Mining Prediction Models and Feature Selection.

Since the irrigated agriculture is the most water-consuming sector in Brazil, it is a challenge to use water in a sustainable way. Evapotranspiration is the combination process of transferring moisture from the earth to the atmosphere by evaporation and transpiration from plants. By estimating this rate of loss, farmers can efficiently manage the crop water requirement and how much water is available. In this work, we propose prediction models, which can estimate the evapotranspiration based on climatic data collected by an automatic meteorological station. Climatic data are multidimensional, therefore by reducing the data dimensionality, then irrelevant, redundant or non-significant data can be removed from the results. In this way, we consider in the proposed solution to apply feature selection techniques before generating the prediction model. Thus, we can estimate the reference evapotranspiration according to the collected climatic variables. The experiments results concluded that models with high accuracy can be generated by M5’ algorithm with feature selection techniques.

Textual Analysis for the Protection of Children and Teenagers in Social Media - Classification of Inappropriate Messages for Children and Teenagers.

Nowadays the Internet is widely used by children and teenagers, where privacy and exposure protection are often not prioritised. This can leave them exposed to paedophiles, who can use a simple chat to start a conversation, which may be the first step towards sexual abuse. In the paper (Falcão Jr. et al, 2016), the authors proposed a tool to detect possible dangerous conversations for a minor in a social network, based on the minors behaviour. However, the proposed tool does not thoroughly address the analyses of the messages exchanged and attempts to detect the suspicious ones in a chat conversation using a superficial approach. This project aims to extend (Falcão Jr. et al, 2016) by automatically classifying the messages exchanged between a minor and an adult in a social network, hence to separate the ones that seem to come from a paedophile from those that seem to be a normal conversation. An experiment with a real conversation was done to test the effectiveness of the created model.

On computing travel time functions from Trajectory Data Streams

Collecting huge volumes of trajectories opens up new opportunities to capture time-varying and uncertain travel costs to traverse segments on a network. This kind of analyses happens to be conducted offline, by means of data mining analysis on historical data. However, there is a need to deal with the incremental nature of spatio-temporal data and maintain the travel time estimation functions by regarding the dynamic behavior of the traffic. In this work, we tackle the problem of creating and maintaining travel time estimation functions by means of trajectory data streams. We propose a new scheme for computing temporal functions using a regression tree with a transition function -- which yield updates in the binary tree by keeping the differentiability of the temporal functions. In the experimental evaluation, which was conducted on real-world dataset, we show the validity of our approach in terms of quality of results and performance.

CUTiS: optimized online ClUstering of Trajectory data Stream

Recent approaches for online clustering of moving objects location are restricted to instantaneous positions. Subse-quently, they fail to capture the behavior of moving objects over time. By continuously tracking sub-trajectories of moving object at each time window, it becomes possible to gain insight on the current behavior and potentially detect mobility patterns in real time. In our previous work [1], we proposed CUTiS, an incremental algorithm for discovering and maintaining the density-based clusters in trajectory data streams, while tracking the evolution of the clusters. This paper extends [1] to CUTiS* by proposing an indexing structure for sub-trajectory data based on a space-filling curve. The proposed index improves the performance of our approach without losing quality in the clusters results as we show in our experiments conducted on a real dataset.