Explore chapters and articles related to this topic
On Error Measures for Validation and Uncertainty Estimation of Predictive QSAR Models
Published in Agnieszka Gajewicz, Tomasz Puzyn, Computational Nanotoxicology, 2019
Supratik Kar, Kunal Roy, Jerzy Leszczynski
A conformal predictor is a type of confidence predictor. The prediction regions generated by a conformal predictor have the property of always being valid, and one needs to worry about their efficiency only. To construct a conformal predictor’s prediction regions, at first the nonconformity measure needs to be defined. Intuitively, this is a way of measuring how different a new example is from old examples. Therefore, a nonconformity measure serves a role similar to that of an AD measure approach. Interestingly, the majority of AD approaches can be employed as nonconformity measures. Conformal prediction is an algorithm-independent technique that works with any predictive method, such as a support vector machine and a random forest (RF), that outputs confidence regions (CRs) for individual predictions in the case of regression and p values for categories in a classification setting [39, 45].
Predicting Amazon customer reviews with deep confidence using deep learning and conformal prediction
Published in Journal of Management Analytics, 2022
In this investigation, we have shown that the combination of deep learning and conformal prediction results in highly predictive and efficient temporal test set sentiment estimates for categories of product reviews with sufficient average word count (∼100 words or more) where the error rate from the derived models can be controlled. For categories of product reviews with insufficient average word count, the present work indicates that cross-category predictions, i.e. using a model of another category for predictions in combination with the calibration set from the investigated review category, also seem to be highly predictive and efficient with only minor loss in terms of efficiency indicating generalizability across product categories. The investigation also shows that the combination of deep learning and conformal prediction gracefully handles class imbalances without explicit class balancing measures. Thus, the mathematical rigor and performance of Mondrian conformal prediction shown in this work makes the framework, agnostic to the machine learning technique employed, an attractive alternative to more traditional approaches for sentiment analysis and other similar tasks.
Analysing and forecasting the security in supply-demand management of Chinese forestry enterprises by linear weighted method and artificial neural network
Published in Enterprise Information Systems, 2021
Machine learning is an inter-discipline that involves multiple fields, in which ANN has attracted wide attention. In the past decade, researches on ANN are numerous. For example, based on the radial basis function neural network, Ahmed et al. (2009) constructed a model for water quality prediction; the results showed that the proposed model was superior to the linear regression model, with an error rate of only 8.3% (Ahmed, Elshafie, and Karim et al. 2009). Guresen et al. (2011) built a hybrid neural network to predict the stock market index (Guresen, Kayakutlu, and Daim 2011). Cireşan et al. (2012) applied the deep neural network (DNN) to realise, as well as improve, the performance of the German Giti tire marks recognition (Cireşan et al. 2012). Wu et al. (2015) proposed a novel credibility guidance method with ANN (Wu, Zhu, and Li. 2015). Currently, researches on ANN have moved a giant step forward. Based on machine learning, Yao et al. (2019) proposed a new calculation method for the absorption boundary condition of the finite difference time domain (FDTD). Compared with traditional methods, this method, combined with ANN, greatly reduced the computational domain size and the computational complexity of FDTD (Yao and Jiang 2019). Peng et al. (2019) applied deep learning to modulation classification of major tasks in communication systems; they adopted two convolutional neural networks (CNNs) in deep learning, i.e. AlexNet and GoogleNet, and compared them with the traditional algorithms (Peng et al. 2019). Jo et al. (2019) utilised CNN to process map images showing traffic states; then, they predicted traffic speed through the input and output images (Jo et al. 2019). Based on 3D CNN, Amyar et al. (2019) proposed an end-to-end prediction model, i.e. 3D RPET-NET. This model extracted 3D image features through four layers, which could improve accuracy (Amyar et al. 2019). Papadopoulos et al. (2009) applied machine learning to medical decision support; the results showed that conformal prediction based on neural network classifiers could be applied to accurately diagnosis acute abdominal pain (Papadopoulos, Gammerman, and Vovk 2009). Claveria et al. (2016) evaluated the influence of prediction level on prediction performance of several machine learning technologies, compared the accuracy of support vector regression (SVR) and neural networks, and improved the prediction accuracy with the linear model as the criterion (Claveria, Monte, and Torra 2016).