Feature engineering for semantic place prediction

Zhu, Yin; Zhong, Erheng; Lu, Zhongqi; Yang, Qiang

doi:10.1016/j.pmcj.2013.07.004

Cited by 29 publications

(16 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The classification results using logistic regression, SVM with different kernels, Gradient Boosted Trees (GBT), and random forests are reported. The authors of [6] have published an extension to paper [21].…”

Section: Discussionmentioning

confidence: 99%

“…Other works have been carried out with similar goals to ours [6][7][8][9]21], that is, semantic place prediction, and use data derived from the same database as data set #1 in our work. They differ from our work in the following aspects: the number of features we used for our classification method is only 14 at most, while the other works use more features; we use different classifiers; while the other papers classify all the 10 labels available in data set #1, we prioritize recognizing Home and Work and therefore combine all the less frequent labels to one label Other; and we present a comparison between three different data representation schemes: visits, places, and cumulative samples.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Semantic Labeling of User Location Context Based on Phone Usage Features

Leppäkoski

Rivero-Rodriguez

Rautalin

et al. 2017

Mobile Information Systems

View full text Add to dashboard Cite

In mobile phones, the awareness of the user's context allows services better tailored to the user's needs. We propose a machine learning based method for semantic labeling that utilizes phone usage features to detect the user's home, work, and other visited places. For place detection, we compare seven different classification methods. We organize the phone usage data based on periods of uninterrupted time that the user has been in a certain place. We consider three approaches to represent this data: visits, places, and cumulative samples. Our main contribution is semantic place labeling using a small set of privacy-preserving features and novel data representations suitable for resource constrained mobile devices. The contributions include (1) introduction of novel data representations including accumulation and averaging of the usage, (2) analysis of the effect of the data accumulation time on the accuracy of the place classification, (3) analysis of the confidence on the classification outcome, and (4) identification of the most relevant features obtained through feature selection methods. With a small set of privacy-preserving features and our data representations, we detect the user's home and work with probability of 90% or better, and in 3-class problem the overall classification accuracy was 89% or better.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Semantic Labeling of User Location Context Based on Phone Usage Features

Leppäkoski

Rivero-Rodriguez

Rautalin

et al. 2017

Mobile Information Systems

View full text Add to dashboard Cite

show abstract

“…In almost all machine-learning tasks, feature engineering and feature selection play an important part [49]. Even if sophisticated estimation algorithms are developed and powerful computing capabilities are available, significant time and energy are spent on looking over the data itself with the goal of identifying additional information, which may be in the Most problems in an M2M domain might be approached at least from two different angles: from the point of view of learning from unlabeled data and from the point of view of active learning.…”

Section: Research Challengesmentioning

confidence: 99%

Big Data Analysis for M2M Networks:Research Challenges and Open Research Issues

Tuna¹,

Daş²,

Ramakrishnan³

et al. 2017

ijcna

View full text Add to dashboard Cite

-In recent years, solutions based on machine-tomachine (M2M) communications have started to support us in many areas of our life and work. However, the amount of data collected by M2M has increased tremendously and surpassed our expectations. This makes it necessary to investigate data mining methodologies and machine learning techniques in order to efficiently utilize large amounts of data gathered by M2M devices. In this paper, we first review existing data mining and machinelearning techniques specifically designed and proposed for M2M networks. Then, we discuss Big Data concept, investigate Big Data analysis techniques, and the importance of Big Data for M2M networks. Finally, we investigate research challenges and open research issues in M2M to provide an insight into future research opportunities.

show abstract

“…Although the problem of automatic location semantics prediction has been discussed in many literatures [7] [14] [18], previous studies usually applied common classification techniques directly, neglecting the fact that such a problem suffers from serious class imbalance [10] [11], e.g., people go to office much often then they go to cinemas. To address the class imbalance problem, we have tried to tackle the automatic location semantics prediction problem by learning a MultiLevel Classification Model in [7], which divides original classification problem into several sub-problems for classification.…”

Section: Introductionmentioning

confidence: 99%

Location semantics prediction for living analytics by mining smartphone data

Huang

Ying

Tseng

et al. 2014

2014 International Conference on Data Science and Advanced Analytics (DSAA)

View full text Add to dashboard Cite

Automatic location semantics prediction for living analytics based on smartphone data has attracted extensive attention in just recent years. Basically, this task can be formulated as a multi-class classification problem, where different location/places are regarded as different labels. Previous studies were mostly based on common classification techniques directly, neglecting the critical challenging issue of class imbalance in such a problem (e.g., people go to offices much more often than they go to cinemas). It is also noteworthy that in contrast to common multi-class problems where the classes can be treated independently and interchangeably, the places for labeling usually have important correlations, which should be taken account in the classification/labeling process. Moreover, several activities may occur in the same place and thus the same place label might convey different semantics. In this paper, we address the above issues for location semantics prediction by proposing the FS-Mining (Frame-based Semantics Mining) approach. We treat the raw sensor data in the smartphone as a sequence of short and non-overlapping frames, based on which the user behavior at each place can be characterized and the place semantics can be modeled. To deal with the issues of label relation and class imbalance, a multi-level classification model with class-split and class-merge mechanisms was also developed. An ensemble strategy was also employed to further improve the performance. Experiments on the dataset of Nokia Mobile Data Challenge [1] demonstrate promising performances for the FSMining approach.

show abstract

Feature engineering for semantic place prediction

Cited by 29 publications

References 21 publications

Semantic Labeling of User Location Context Based on Phone Usage Features

Semantic Labeling of User Location Context Based on Phone Usage Features

Big Data Analysis for M2M Networks:Research Challenges and Open Research Issues

Location semantics prediction for living analytics by mining smartphone data

Contact Info

Product

Resources

About