Hisashi Kurasawa

Recognizing User Context Using Mobile Handsets with Acceleration Sensors

User context recognition is one of the important technologies for realizing context aware services. Conventional multi sensor based approach has advantages in that it can generate variety of contexts with less computation resources by using many different sensors. However, such systems tend to be complex and cumbersome and, thus, do not fit in well with mobile environment. In this sense, a single sensor based approach is suitable for mobile environments. In this paper, we show a context inference scheme that realizes a user posture inference with only one acceleration sensor embedded in a mobile handset. Our system automatically detects the sensor position on the users body and selects the most relevant inference method dynamically. Our experimental results show that the system can infer a users posture (sitting, standing, walking, and running) with an accuracy of more than 96%.

Top of worlds: method for improving motivation to participate in sensing services

We propose a method for improving motivation to participate in sensing services by presenting rankings in multidimensional hierarchical sets. We call this method Top of Worlds. Because previously proposed methods only rank a user among all other users, many have little chance of being ranked in the top group, resulting in little motivation to continue. Top of Worlds creates many sets with varying granularity to increase the chance of many users being ranked in the top group and presents these rankings in those sets. Through an experiment, we partially confirmed the validity of Top of Worlds.

Finding the k-closest pairs in metric spaces

We investigated the problem of reducing the cost of searching for the k closest pairs in metric spaces. In general, a k-closest pair search method initializes the upper bound distance between the k closest pairs as infinity and repeatedly updates the upper bound distance whenever it finds pairs of objects whose distances are shorter than that distance. Furthermore, it prunes dissimilar pairs whose distances are estimated as longer than the upper bound distance based on the distances from the pivot to objects and the triangle inequality. The cost of a k-closest pair query is smaller for a shorter upper bound distance and a sparser distribution of distances between the pivot and objects. We propose a new divide-and-conquer-based k-closest pair search method in metric spaces, called Adaptive Multi-Partitioning (AMP). AMP repeatedly divides and conquers objects from the sparser distance-distribution space and speeds up the convergence of the upper bound distance before partitioning the denser space. As a result, AMP can prune many dissimilar pairs compared with ordinary divide-and-conquer-based method. We compare our method with other partitioning method and show that AMP reduces distances computations.

Maximal metric margin partitioning for similarity search indexes

We propose a partitioning scheme for similarity search indexes that is called Maximal Metric Margin Partitioning (MMMP). MMMP divides the data on the basis of its distribution pattern, especially for the boundaries of clusters. A partitioning surface created by MMMP is likely to be at maximum distances from the two cluster boundaries. MMMP is the first similarity search index approach to focus on partitioning surfaces and data distribution patterns. We also present an indexing scheme, named the MMMP-Index, which uses MMMP and small ball partitioning. The MMMP-Index prunes many objects that are not relevant to a query, and it reduces the query execution cost. Our experimental results show that MMMP effectively indexes clustered data and reduces the search cost. For clustered vector data, the MMMP-Index reduces the computational cost to less than two thirds that of comparable schemes.

Data allocation scheme based on term weight for P2P information retrieval

Many Peer-to-Peer information retrieval systems that use a global index have already been proposed that can retrieve documents relevant to a query. Since documents are allocated to peers regardless of the query, the system needs to connect many peers to gather the relevant documents. We propose a new data allocation scheme for P2P information retrieval that we call Concordia. Concordia uses a node to allocate a document based on the weight of each term in the document to efficiently assemble all the documents relevant to a query from the P2P Network. Moreover, the node encodes the binary data of a document with an erasure code, and Concordia produces an efficient redundancy for counteracting node failures.

Machine-Learning-Based Prediction of a Missed Scheduled Clinical Appointment by Patients With Diabetes.

Background: About 10% of patients with diabetes discontinue treatment, resulting in the progression of diabetes-related complications and reduced quality of life. Objective: The objective was to predict a missed clinical appointment (MA), which can lead to discontinued treatment for diabetes patients. Methods: A machine-learning algorithm was used to build a logistic regression model for MA predictions, with L2-norm regularization used to avoid over-fitting and 10-fold cross validation used to evaluate prediction performance. Data associated with patient MAs were extracted from electronic medical records and classified into two groups: one related to patients’ clinical condition (X1) and the other related to previous findings (X2). The records used were those of the University of Tokyo Hospital, and they included the history of 16 026 clinical appointments scheduled by 879 patients whose initial clinical visit had been made after January 1, 2004, who had diagnostic codes indicating diabetes, and whose HbA1c had been tested within 3 months after their initial visit. Records between April 1, 2011, and June 30, 2014, were inspected for a history of MAs. Results: The best predictor of MAs proved to be X1 + X2 (AUC = 0.958); precision and recall rates were, respectively, 0.757 and 0.659. Among all the appointment data, the day of the week when an appointment was made was most strongly associated with MA predictions (weight = 2.22). Conclusions: Our findings may provide information to help clinicians make timely interventions to avoid MAs.

An evaluation of method for encouraging participation

Much attention is being focused on participatory sensing, in which real-world data are collected using personal mobile devices as sensor nodes to sense various conditions of the world we live in. In participatory sensing, there is a problem in that the supply of data is insufficient if users are not motivated to participate in sensing services. We previously proposed Top of Worlds, a method for encouraging user participation by presenting rankings in multidimensional hierarchical sets. In this paper, we describe the development of a ranking system and a real-world evaluation to confirm that Top of Worlds can encourage user participation.

GPS Trajectory Data Enrichment based on a Latent Statistical Model

This paper proposes a latent statistical model for analyzing global positioning system (GPS) trajectory data. Because of the rapid spread of GPS-equipped devices, numerous GPS trajectories have become available, and they are useful for various location-aware systems. To better utilize GPS data, a number of sensor data mining techniques have been developed. This paper discusses the application of a latent statistical model to two closely related problems, namely, moving mode estimation and interpolation of the GPS observation. The proposed model estimates a latent mode of moving objects and represents moving patterns according to the mode by exploiting a large GPS trajectory dataset. We evaluate the effectiveness of the model through experiments using the GeoLife GPS Trajectories dataset and show that more than three-quarters of covered locations were correctly reproduced by interpolation at a fine granularity.

Huffman-DHT: Index Structure Refinement Scheme for P2P Information Retrieval

Peer-to-peer information retrieval (P2P IR) systems using a distributed index on a distributed hash table (DHT) can make highly precise searches for documents relevant to a query. However, these systems require a heavy index construction cost, and cause unfair index management costs due to the unbalanced term frequency distribution. We propose a new node access scheme for P2P IR that we call Huffman-DHT. Huffman-DHT uses an algorithm similar to Huffman coding, and modifies the DHT structure based on the term distribution. Huffman-DHT distributes the index construction cost among the nodes equally, and achieves load balancing.

Top of worlds: estimating time complexity of calculating rank order in multi-dimensional hierarchical sets

The increasing number of mobile devices, such as smartphones, has brought attention to participatory sensing in which real-world data are collected via personal devices. To collect data via participatory sensing, it is important to motivate participants. Thus, we previously proposed Top of Worlds, a method for encouraging user participation by presenting their rank order. In this paper, we estimate its time complexity in order to understand how often we can present a rank order in planning phase of services.