Sensor-enabled smartphones are opening a new frontier in the development of mobile sensing applications. The recognition human activities and context from sensor-data using classification models underpins these emerging However, conventional approaches to training classifiers struggle cope with diverse user populations routinely found large-scale popular Differences between users (e.g., age, sex, behavioral patterns, lifestyle) confuse classifiers, which assume everyone is same. To address...

Reliable smartphone app prediction can strongly benefit both users and phone system performance alike. However, real-world usage behavior is a complex phenomena driven by number of competing factors. In this pa- per, we develop an model that leverages three key everyday factors affect decisions -- (1) intrinsic user preferences historical patterns; (2) activities the environment as observed through sensor-based contextual signals; and, (3) shared aggregate patterns appear in various...

For decades large corporations as well labor placement services have maintained extensive yet static resume databanks. Online professional networks like LinkedIn taken these databanks to a dynamic, constantly updated and massive scale profile dataset spanning career records from hundreds of industries, millions companies people worldwide. Using this dataset, paper attempts model profiles individuals sequence positions held by them time-series nodes, each which represents one particular...

The Cooperative Communities (CoCo) learning framework leverages everyday social connections between people to personalize classification models. By exploiting networks, CoCo spreads the burden of providing training data over an entire community.

Multi-instance learning, like other machine learning and data mining tasks, requires distance metrics. Although metric methods have been studied for many years, learners multi-instance remain almost untouched. In this paper, we propose a framework called Multi-Instance MEtric Learning (MIMEL) to learn an appropriate under the setting. The between two bags is defined using Mahalanobis function. problem formulated by minimizing KL divergence multivariate Gaussians constraints of maximizing...

Networks that model relationships in the real world have attracted much attention past few years. Link prediction plays a central role network area. Supervised learning is an important class of algorithms used to address link problem. A big challenge solving tasks deciding how choose relevant features. As machine technique select features, feature selection not only enhances classification accuracy, but also improves efficiency training process. Thus, it especially for prediction. However,...

Human behavior modeling at a large-scale and under real-world conditions is still an open problem. Existing classification models do not always perform well on diverse population. Training personalized that incorporate different contextual individual user characteristics are effective in addressing this challenge. However, approach burdens the users with collecting manually labeling their own training data which scalable. In article, we propose CoCo (Cooperative Communities), learning...

This paper proposes another technique to forecast Chinese mobile user based on the ARIMA model. There are two major premises use model in prediction: market is mature and still has big space for development. On basis of historical incremental number from 2002q1 2009q4, through identification estimation, an (1, 1, 2) selected fit time series. After diagnostic evaluation, 2010 quarterly predicted. The approved be accurate short-time prediction.

The single greatest opportunity to improve health and reduce premature death lies in personal behavior. While technology-based behavior intervention has been around for many years, the emerging smartphone wearable sensing technology brings great promise push change further by inferring predicting real-time occurrence its context. In this paper, we envision how social physical context awareness could sustain motivation assist habit formation. We describe our preliminary work that supports...

Previous chapter Next Full AccessProceedings Proceedings of the 2011 SIAM International Conference on Data Mining (SDM)FAMER: Making Multi-Instance Learning Better and FasterWei Ping, Ye Xu, Jianyong Wang, Xian-Sheng HuaWei Huapp.594 - 605Chapter DOI:https://doi.org/10.1137/1.9781611972818.51PDFBibTexSections ToolsAdd to favoritesExport CitationTrack CitationsEmail SectionsAboutAbstract Kernel method is a powerful tool in multi-instance learning. However, many typical kernel methods for...

In wireless sensor networks, the nodes usually use battery power, one of main design challenges is to obtain long system lifetime. a network, sense several kinds information. These data have different characteristics. this paper, sensed from are divided into two types; they emergent and usual data. And fusion layer proposed. By using reduction algorithm, only transfer changed information sink node. node recruits full Finally, we implement algorithm in TinyOS. Experiment results show that can...

For the modeling problem of microbial fermentation process, taking glutamic acid process as research object, decision tree and random forest model were established by using data mining method, was evaluated predicted R language. Good effect model, indicating that package language has a certain flexibility, through choice parameters can be useful model. In addition, under same conditions, is constructed. The simulation results show combined based on algorithm superior to single prediction...

A scale-free routing protocol and algorithm in wireless sensor networks are studied this paper. The could be used to form a WSN with scale-free-network-like topology which helps shorten the length of path. Together information fusion function involved protocol, dramatically reduce redundancies datagram transferred thus improve life cycle WSN.

Feature selection is an effective tool to deal with the "curse of dimensionality". To cope non-separable problem, feature in kernel space has been investigated. However, previous study cannot adequately estimate intrinsic dimensionality space. Thus, it difficult accurately preserve sketch using learned basis, and performance affected. Moreover, computing load algorithm reaches at least cubic number training data. In this paper, we propose a fast framework conduct By designing subspace...

The condensed cube has been proposed to reduce the huge size of data cubes in OLAP system. intuition is compress semantically redundant tuples into their representative base single (BSTs). However, previous studies showed that a minimal expensive compute, and thus mainly concentrated on alternative computation methods for non-minimal cube, which does not guarantee find all BSTs. In this paper, we focus address several practical issues, including physical organization, fast computation,...

To evaluate how much two different complex topologies are similar to each other in a quantitative way is an essential procedure large‐scale topology researches and still remains NP problem. Cross‐correlation evaluation model (CCEM) together with Genetic Algorithm (GA) introduced this paper trying solve issue. Experiments have proved that SLS (Signless Laplacian Spectra) capable of identifying structure CCEM distinguishing the differences between corresponding eigenvectors. used GA...

A fault detection system of a power plant by means neuro-fuzzy fusion method is discussed in this paper. Stator temperature together with variations refrigerant at the entrance and exit hydrogenerator group are firstly monitored sensors. The fuzzy aforementioned variables their membership functions, next, designed according to expertise knowledge base. Finally, neuron-fuzzy model generated merge into neural networks model, it proved be efficient correctness ratio 91% testing experiments on...