Liping Wang

Spatial Keyword Range Search on Trajectories

With advances in geo-positioning technologies and ensuing location based service, there are a rapid growing amount of trajectories associated with textual information collected in many emerging applications. For instance, nowadays many people are used to sharing interesting experience through Foursquare or Twitter along their travel routes. In this paper, we investigate the problem of spatial keyword range search on trajectories, which is essential to make sense of large amount of trajectory data. To the best of our knowledge, this is the first work to systematically investigate range search over trajectories where three important aspects, i.e., spatio, temporal and textual, are all taken into consideration. Given a query region, a timespan and a set of keywords, we aim to retrieve trajectories that go through this region during query timespan, and contain all the query keywords. To facilitate the range search, a novel index structure called IOC-Tree is proposed based on the inverted indexing and octree techniques to effectively explore the spatio, temporal and textual pruning techniques. Furthermore, this structure can also support the query with order-sensitive keywords. Comprehensive experiments on several real-life datasets are conducted to demonstrate the efficiency.

Multi-lead ECG classification based on Independent Component Analysis and Support Vector Machine

An novel multi-lead Electrocardiogram (ECG) classification method is proposed in this paper. At the feature extracting stage, an improved Independent Component Analysis (ICA) method is introduced. In our method, a heartbeat is intercepted into 3 segments (P wave, QRS interval, ST segment). ICA is used to extract the features of each segment separately. These three feature vectors construct the feature of single lead firstly. Then, twelve single lead feature vectors are combined to generate a multi-lead feature vector one by one. At last, the Support Vector Machine (SVM) is used for multi-classification and 2-classification experiments. All available data in MIT-BIH Arrhythmia Database and the number of 2500 practical data gathered from about 500 persons is used in experiments simultaneously. For MIT-BIH data, multi-classification result is discussed. The final average accuracy of the testing data is 98.18% and the average sensitivity is 98.68%. For practical data, 2-classification experiment result is discussed. The accuracy of testing data is 90.47% and the sensitivity is 90.01%.

Chinese Cardiovascular Disease Database (CCDD) and Its Management Tool

Standard Electrocardiogram (ECG) database is prepared for testing the performance of automatic detection and classification algorithms. At present, there are three mainstream standard databases used by computer-aided ECG diagnosis researchers: MIT-BIH arrhythmia database, CSE multi-lead database and AHA database. By the progress of ECG in both equipment and diagnosis theory, fatal deficiency was found in these databases and a new one is needed for further studies. So Chinese Cardiovascular Disease Database (CCDD or CCD database), which contains 12-Lead ECG data and detailed features with diagnosis result is proposed. It is distinguished not only by improving the raw ECG data’s technical parameters, but also introduces some morphology features. Investigation shows these features are utilized by experienced cardiologists effectively. CCDD is used in our group as well as aiming for other and others’ projects in the future.

GALLOP: GlobAL Feature Fused LOcation Prediction for Different Check-in Scenarios

Location prediction is widely used to forecast users’ next place to visit based on his/her mobility logs. It is an essential problem in location data processing, invaluable for surveillance, business, and personal applications. It is very challenging due to the sparsity issues of check-in data. An often ignored problem in recent studies is the variety across different check-in scenarios, which is becoming more urgent due to the increasing availability of more location check-in applications. In this paper, we propose a new feature fusion based prediction approach, GALLOP, i.e., GlobAL feature fused LOcation Prediction for different check-in scenarios. Based on the carefully designed feature extraction methods, we utilize a novel combined prediction framework. Specifically, we set out to utilize the density estimation model to profile geographical features, i.e., context information, the factorization method to extract collaborative information, and a graph structure to extract location transition patterns of users’ temporal check-in sequence, i.e., content information. An empirical study on three different check-in datasets demonstrates impressive robustness and improvement of the proposed approach.

Efficiently Retrieving Top-k Trajectories by Locations via Traveling Time

The flourishing industry of location-based services has collected a massive amount of users’ positions in the form of spatial trajectories, which raise many research problems. In this paper, we study a trajectory retrieving query, k-TLT, which aims at retrieving the top-k Trajectories by Locations and ranked by traveling Time. Given a set Q of query locations, a k-TLT query retrieves top-k trajectories that are close to Q with respect to traveling time. In contrast to existing works which consider only location information, k-TLT queries also consider the traveling time information, which have many applications, such as travel route planning and moving object study. To efficiently answer a k-TLT query, we first online compute a list L q of trajectories for each query location q ∈ Q, such that trajectories in L q are ranked by their traveling time to q. Based on the online generated lists L q corresponding to query locations, a small set of candidate trajectories that are close to Q is selected by iteratively retrieving trajectories from lists L q . Then, the set of candidate trajectories is refined and pruned to determine the top-k trajectories. We conduct extensive experiments on a real trajectory dataset and verify the efficiency of our approach.

A time-series similarity method for QRS morphology variation analysis

Electrocardiography is a common tool for detecting cardiovascular system diseases. In clinical, as the individual difference is an intrinsic feature of ECG, data distribution difference between training and testing data impacts on the accuracy of classifier. Automatic ECG classification satisfied clinical demand is urgently required. QRS is a main waves in a heartbeat. In this paper, we propose a complete framework for individual oriented QRS morphology variation analysis. The original signal is first preprocessed by re-sampling and smoothing, then symbolized by dynamic and static combined method. For similarity measure, an improved information entropy measure function based on the symbolic result is proposed and ECG domain knowledge is well utilized by the function. At last, the entropy function based unsupervised learning algorithm is presented for QRS complex similarity computation. Our algorithm dedicates to the individual data analysis combined with domain knowledge, which is free from any training data and more suitable for application. Comprehensive experiments show that the proposed entropy function achieves improvements over the general distance measure functions during QRS similarity measure. The clustering algorithm is effective at recognizing normal and abnormal QRS morphology.

Comprehensive Graph and Content Feature Based User Profiling

Nowadays, users post a lot of their ordinary life records to online social sites. Rich social content covers discussion, interaction and communication activities etc. The social data provides insights into users’ interest, preference and communication aspects. An interesting problem is how to profile users’ occupation, i.e., professional categories. It has great values for users’ recommendation and personalized delivery services. However, it is very challenging, compared to gender or age prediction, due to the multiple categories and complex scenarios.

A Peer to Peer Housing Rental System with Continuous Spatial-Keyword Searching

Apart from traditional intermediary companies, online housing rental systems as their convenience are gaining their popularity. Search is a critical function in these systems, but it does not always meet users’ satisfaction. For example, spatial and keyword distributions are generally be considered separately and the user cannot submit a continuous query requirement before he/she rent a satisfied house. In this paper, we develop a peer to peer housing rental system (P2PHRS) based on Django Framework of Python. We propose an efficient house searching algorithm called Quad-tree plus Inverted List (QIL) to filter housing resources for users according to their spatial and keyword requirements. P2PHRS is designed to be adaptive to a variety of front-end clients, like Web, Android and iOS platform etc. We show the advantages of P2PHRS by several spatial-keyword query demonstration scenarios.

Encoding Document Semantic into Binary Codes Space

We develop a deep neural network model to encode document semantic into compact binary codes with the elegant property that semantically similar documents have similar embedding codes. The deep learning model is constructed with three stacked auto-encoders. The input of the lowest auto-encoder is the representation of word-count vector of a document, while the learned hidden features of the deepest auto-encoder are thresholded to be binary codes to represent the document semantic. Retrieving similar document is very efficient by simply returning the documents whose codes have small Hamming distances to that of the query document. We illustrate the effectiveness of our model on two public real datasets – 20NewsGroup and Wikipedia, and the experiments demonstrate that the compact binary codes sufficiently embed the semantic of documents and bring improvement in retrieval accuracy.

Efficient processing node proximity via random walk with restart

Graph is a useful tool to model complicated data structures. One important task in graph analysis is assessing node proximity based on graph topology. Recently, Random Walk with Restart (RWR) tends to pop up as a promising measure of node proximity, due to its proliferative applications in e.g. recommender systems, and image segmentation. However, the best-known algorithm for computing RWR resorts to a large LU matrix factorization on an entire graph, which is cost-inhibitive. In this paper, we propose hybrid techniques to efficiently compute RWR. First, a novel divide-and-conquer paradigm is designed, aiming to convert the large LU decomposition into small triangular matrix operations recursively on several partitioned subgraphs. Then, on every subgraph, a “sparse accelerator” is devised to further reduce the time of RWR without any sacrifice in accuracy. Our experimental results on real and synthetic datasets show that our approach outperforms the baseline algorithms by at least one constant factor without loss of exactness.