Optimizing the Multiclass F-Measure via Biconcave Programming

Narasimhan, Harikrishna; Pan, Weiwei; Kar, Purushottam; Protopapas, Pavlos; Ramaswamy, Harish G.

doi:10.1109/icdm.2016.0143

Cited by 26 publications

(16 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the classification problem, class imbalance is common because most datasets do not contain exactly equal number of instances in each class, and this phenomenon also applies to human activities during daily lives. F1-score is a commonly used measure in the class imbalanced settings [41]. In the multi-class classification domain, however, micro-averaging F1-score is equivalent to accuracy which measures a ratio of correctly predicted observation to the total instances.…”

Section: Discussionmentioning

confidence: 99%

Sensor Data Acquisition and Multimodal Sensor Fusion for Human Activity Recognition Using Deep Learning

Chung

Lim

Noh

et al. 2019

Sensors

147

View full text Add to dashboard Cite

In this paper, we perform a systematic study about the on-body sensor positioning and data acquisition details for Human Activity Recognition (HAR) systems. We build a testbed that consists of eight body-worn Inertial Measurement Units (IMU) sensors and an Android mobile device for activity data collection. We develop a Long Short-Term Memory (LSTM) network framework to support training of a deep learning model on human activity data, which is acquired in both real-world and controlled environments. From the experiment results, we identify that activity data with sampling rate as low as 10 Hz from four sensors at both sides of wrists, right ankle, and waist is sufficient in recognizing Activities of Daily Living (ADLs) including eating and driving activity. We adopt a two-level ensemble model to combine class-probabilities of multiple sensor modalities, and demonstrate that a classifier-level sensor fusion technique can improve the classification performance. By analyzing the accuracy of each sensor on different types of activity, we elaborate custom weights for multimodal sensor fusion that reflect the characteristic of individual activities.

show abstract

Section: Discussionmentioning

confidence: 99%

Sensor Data Acquisition and Multimodal Sensor Fusion for Human Activity Recognition Using Deep Learning

Chung

Lim

Noh

et al. 2019

Sensors

147

View full text Add to dashboard Cite

show abstract

“…Sensitivity, specificity and predictive accuracy of the decision algorithms (A) and (B) were calculated. As decision tree C is a multiclass predictor, sensitivity and specificity were not directly defined for it, and the corresponding mean sensitivity, mean specificity, and mean accuracy were calculated as previously described [19]. These quality measures were obtained as follows: For every multiclass predictor, which differentiates between n classes, n binary sub-predictors were defined.…”

Section: Discussionmentioning

confidence: 99%

Novel decision algorithm to discriminate parkinsonism with combined blood and imaging biomarkers

Mangesius

Mariotto

Ferrari

et al. 2020

Parkinsonism & Related Disorders

View full text Add to dashboard Cite

Introduction: To determine an exploratory multimodal approach including serum NFL and MR planimetric measures to discriminate Parkinson's disease (PD), multiple system atrophy (MSA) and progressive supranuclear palsy (PSP). Methods: MR planimetric measurements and NFL serum levels, with a mean time interval of 60 months relative to symptom onset, were assessed in a retrospective cohort of 11 progressive supranuclear palsy (PSP), 22 Parkinson's disease (PD), 16 multiple system atrophy (MSA) patients and 42 healthy controls (HC). A decision tree model to discriminate PD, PSP, and MSA was constructed using receiver operating characteristic curve analysis and Classification and Regression Trees algorithm. Results: Our multimodal decision tree provided accurate differentiation of PD versus MSA and PSP patients using a serum NFL cutoff of 14.66 ng/L. The pontine-to-midbrain-diameter-ratio (P d /M d) discriminated MSA from PSP at a cutoff value of 2.06. The combined overall diagnostic yield was an accuracy of 83.7% (95% CI 69.8-90.8%). Conclusion: We provide a clinically feasible decision algorithm which combines serum NFL levels and a planimetric MRI marker to differentiate PD, MSA and PSP with high diagnostic accuracy. Classification of evidence: This study provides Class III evidence that the combination of serum NFL levels und MR planimetric measurements discriminates between PD, PSP and MSA.

show abstract

“…For simplicity in evaluation, we calculate metrics for each class and only report the macro-averaged values over all classes. We prefer macro average in unbalanced settings because it calculate the unweighted mean and has equal emphasis on all classes [45]. When only one metric is considered for evaluation, we use F1 score by default.…”

Section: ) Evaluation Metricsmentioning

confidence: 99%

Automatic Mobile App Identification From Encrypted Traffic With Hybrid Neural Networks

Wang

Chen

2020

IEEE Access

View full text Add to dashboard Cite

The proliferation of handheld devices has led to an explosive growth of mobile traffic volumes on the Internet. Network traffic classification in mobile settings, also known as mobile app identification, has become increasingly important. It is an essential step to enable network management, security operations and other services. Traditionally, the design of accurate identifiers relies on the deep packet inspection (DPI) techniques. However, such approaches have become less effective with the raising adoption of encrypted protocols in mobile applications (mostly TLS). To address the problem, various machine learning methods have been studied and used. Most of them use linear classifiers on top of hand-engineered features, which are unreliable due to the complexity of mobile traffic. In this paper we propose App-Net, an end-to-end hybrid neural network for mobile app identification from encrypted TLS traffic. App-Net is designed by combining RNN and CNN in a parallel way and can automatically learn effective features from raw TLS flows. With coordinated fusion and optimized training, the hybrid and multimodal architecture is able to characterize both flow sequence patterns and app signatures to learn a joint flow-app embedding. We evaluate App-Net on a real-world dataset covering 80 apps. The results show that our method can achieve an excellent performance and outperform the state-of-the-art methods. INDEX TERMS Mobile app identification, encrypted traffic classification, neural network, deep learning.

show abstract

Optimizing the Multiclass F-Measure via Biconcave Programming

Cited by 26 publications

References 6 publications

Sensor Data Acquisition and Multimodal Sensor Fusion for Human Activity Recognition Using Deep Learning

Sensor Data Acquisition and Multimodal Sensor Fusion for Human Activity Recognition Using Deep Learning

Novel decision algorithm to discriminate parkinsonism with combined blood and imaging biomarkers

Automatic Mobile App Identification From Encrypted Traffic With Hybrid Neural Networks

Contact Info

Product

Resources

About