Optimal feature selection for classifying a large set of chemicals using metal oxide sensors

Nowotny, Thomas; Berna, Amalia Z.; Binions, Russell; Trowell, Stephen

doi:10.1016/j.snb.2013.01.088

Cited by 36 publications

(73 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(2) For feature sets of small size constraints, Sensor 8 was utilised over 50% of the times in [7], while here Sensor 10 was picked more than 80% of the times. Further, Sensor 1 and 6's appearances for size constraint of 1 are swapped.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Feature Selection for Chemical Sensor Arrays Using Mutual Information

et al. 2014

Self Cite

View full text Add to dashboard Cite

We address the problem of feature selection for classifying a diverse set of chemicals using an array of metal oxide sensors. Our aim is to evaluate a filter approach to feature selection with reference to previous work, which used a wrapper approach on the same data set, and established best features and upper bounds on classification performance. We selected feature sets that exhibit the maximal mutual information with the identity of the chemicals. The selected features closely match those found to perform well in the previous study using a wrapper approach to conduct an exhaustive search of all permitted feature combinations. By comparing the classification performance of support vector machines (using features selected by mutual information) with the performance observed in the previous study, we found that while our approach does not always give the maximum possible classification performance, it always selects features that achieve classification performance approaching the optimum obtained by exhaustive search. We performed further classification using the selected feature set with some common classifiers and found that, for the selected features, Bayesian Networks gave the best performance. Finally, we compared the observed classification performances with the performance of classifiers using randomly selected features. We found that the selected features consistently outperformed randomly selected features for all tested classifiers. The mutual information filter approach is therefore a computationally efficient method for selecting near optimal features for chemical sensor arrays.

show abstract

Section: Resultsmentioning

confidence: 99%

“…The results in Figure 7 were obtained using linear SVM with cost parameter , which is the same as the method used in [7]. Comparing the two sets of results we can see that there are slight differences in performance.…”

Section: Resultsmentioning

confidence: 99%

Feature Selection for Chemical Sensor Arrays Using Mutual Information

et al. 2014

Self Cite

View full text Add to dashboard Cite

show abstract

“…This is because the optimal number of sensors is a balance between providing sufficient information and redundancy while minimizing the increased complexity, noise, and risk of overfitting associated with high-dimensional spaces. Thus, including more sensors can potentially decrease performance, as has been experimentally observed (138). Such issues are well-known in the areas of machine learning, data mining, and pattern recognition, where so-called curse of dimensionality refers to phenomena that arise out of exponential scaling of measurement space volume with the dimensionality of a sensor array.…”

Section: General-purpose Array Design Considerationsmentioning

confidence: 99%

Sensor Array Design for Complex Sensing Tasks

Johnson

Rose-Pehrsson²

2015

Annual Rev. Anal. Chem.

View full text Add to dashboard Cite

Chemical detection in complex environments presents numerous challenges for successful implementation. Arrays of sensors are often implemented for complex chemical sensing tasks, but systematic understanding of how individual sensor response characteristics contribute overall detection system performance remains elusive, with generalized strategies for design and optimization of these arrays rarely reported and even less commonly adopted by practitioners. This review focuses on the literature of nonspecific sensor array design and optimization strategies as well as related work that may inform future efforts in complex sensing with arrays.

show abstract

“…It should be noted that clear separation of clusters does not necessarily lead to high classification performance. 27 To evaluate further the applicability of the pattern recognition of solid materials, we demonstrated the identification of the molecular weights of polymers. Two additional polystyrenes with different molecular weights, PS(35k) and PS(280k), were also coated onto separate MSS channels in the same manner.…”

Section: Materials Horizonsmentioning

confidence: 99%