Advances in computational intelligence最新文献

Cash-futures basis forecasting represents a vital concern for various market participants in the agricultural sector, which has been rarely explored due to limitations on data and traditional econometric methods. The current study explores usefulness of the nonlinear autoregressive neural network technique for the forecasting problem in a unique and proprietary data set of daily corn cash-futures basis across nearly five-hundred cash markets from sixteen most important harvest states in the United States over a 5-year period. Through investigations of various model settings across the hidden neuron, delay, data splitting ratio, and algorithm, a chosen model with five delays and twenty hidden neurons is reached, trained using the Levenberg–Marquardt algorithm and data splitting ratio of 70% vs. 15% vs. 15% for training, validation, and testing. This model results in accurate and stable performance across the cash markets explored, which illustrates usefulness of the machine learning technique for corn cash-futures basis forecasting. Particularly, the model leads to average relative root mean square errors (RRMSEs) of 9.97%, 8.51%, and 9.64% for the training, validation, and testing phases, respectively, and the average RRMSE of 9.83% for the overall sample across all cash markets. Results here might be used as standalone technical forecasts or combined with fundamental forecasts for forming perspectives of cash-futures basis trends and carrying out policy analysis. The empirical framework here is easy to implement, which is an essential consideration to many decision makers, and has potential to be generalized for forecasting cash-futures basis of other commodities.

Model optimization in deep learning (DL) and neural networks is concerned about how and why the model can be successfully trained towards one or more objective functions. The evolutionary learning or training process continuously considers the dynamic parameters of the model. Many researchers propose a deep learning-based solution by randomly selecting a single classifier model architecture. Such approaches generally overlook the hidden and complex nature of the model’s internal working, producing biased results. Larger and deeper NN models bring many complexities and logistic challenges while building and deploying them. To obtain high-quality performance results, an optimal model generally depends on the appropriate architectural settings, such as the number of hidden layers and the number of neurons at each layer. A challenging and time-consuming task is to select and test various combinations of these settings manually. This paper presents an extensive empirical analysis of various deep learning algorithms trained recursively using permutated settings to establish benchmarks and find an optimal model. The paper analyzed the Stack Overflow dataset to predict the quality of posted questions. The extensive empirical analysis revealed that some famous deep learning algorithms such as CNN are the least effective algorithm in solving this problem compared to multilayer perceptron (MLP), which provides efficient computing and the best results in terms of prediction accuracy. The analysis also shows that manipulating the number of neurons alone at each layer in a network does not influence model optimization. This paper’s findings will help to recognize the fact that future models should be built by considering a vast range of model architectural settings for an optimal solution.

Determining the side effects of multidrug combinations is a very important issue in drug risk studies. However, designing clinical trials to determine frequencies is often time-consuming and expensive, and previous work has often been limited to using the target protein of a drug without screening. Although this alleviates the sparsity of the raw data to some extent, blindly introducing proteins as auxiliary information can lead to a large amount of noisy information being added, which in turn leads to less efficient models. For this reason, we propose a new method called Gorge (graph convolutional networks on heterogeneous multi-relational graphs for polypharmacy side effect prediction). Specifically, we designed two protein auxiliary pathways directly related to drugs and combined these two auxiliary pathways with a multi-relational graph of drug side effects, which both alleviates data sparsity and filters noisy data. Then, we introduce a query-aware attention mechanism that generates different attention pathways for drug entities based on different drug pairs, fine-grained to determine the extent of information delivery. Finally, we output the exact frequency of drug side effects occurring through a tensor factorization decoder, in contrast to most existing methods that can only predict the presence or association of side effects, but not their frequency. We found that Gorge achieves excellent performance on real-world datasets (average AUROC of 0.822 and average AUPR of 0.775), outperforming existing methods. Further examination provides literature evidence for highly ranked predictions.

Diabetes is nowadays a very common medical problem among the people worldwide. The disease is becoming more prevalent with the modern and hectic lifestyle followed by people. As a result, designing an adequate medical expert system to assist physicians in treating the disease on time is critical. Expert systems are required to identify the major cause(s) of the disease, so that precautionary measures can be taken ahead of time. Several medical expert systems have already been proposed, but each has its own set of shortcomings, such as the use of trial and error methods, trivial decision-making procedures, and so on. As a result, this paper proposes a Transparent Diabetes Management System Using Machine Learning (TDMSML) expert system that uses decision tree rules to identify the major factor(s) of diabetes. The TDMSML model comprises of three phases: rule generation, transparent rule selection, and major factor identification. The rule generation phase generates rules using decision tree. Transparent rule selection stage selects the transparent rules followed by pruning the redundant rules to get the minimized rule-set. The major factor identification stage extracts the major factor(s) with range(s) from the minimized rule-set. These factor(s) with certain range(s) are characterized as major cause(s) of diabetes disease. The model is validated with the Pima Indian diabetes data set collected from Kaggle.

The paper explores advancements in accounting analytics using the logical analysis of data (LAD) approach by identifying fraudulent firms and transactions. The straightforward approach to fraud detection with an analytic model is to identify possible predictors of fraud associated with known fraudsters and their historical actions. LAD is a machine learning methodology that combines Boolean functions, optimization, and logic ideas in alignment with the traditional approach. The key characteristic of the LAD is discovering minimal sets of features necessary for explaining all observations and detecting hidden patterns in the data capable of distinguishing observations describing “positive” outcome events from “negative” outcome events. The combinatorial optimization model described in the paper represents a variation on the general theme of set covering and concludes with an outline of LAD applications to detect fraudulent firms and financial frauds. The dataset consists of Annual data of 777 firms from 14 different sectors. The results demonstrate 97.4% accuracy with an F1 score of 0.97. Another dataset on credit card transactions and finance is also used to test the effectiveness of LAD in finance. With the appearance of the immense growth of financial fraud cases, these promising results lead to future advancements in analytical audit fieldwork.

The recommender system is a set of data recovery tools and techniques used to recommend items to users based on their selection. To improve the accuracy of the recommendation, the use of additional information (e.g., social information, trust, item tags, etc.) in addition to user-item ranking data has been an active area of research for the past decade.

In this paper, we present a new method for recommending top-N items, which uses structural information and trust among users within the social network and extracts the implicit connections between users and uses them in the item recommendation process. The proposed method has seven main steps: (1) extract items liked by neighbors, (ii) constructing item features for neighbors, (iii) extract embedding trust features for neighbors, (iv) create user-feature matrix, (v) calculate user’s priority, (vi) calculate item’s priority and finally, (vii) recommend top-N items. We implement the proposed method with three datasets for recommendations. We compare our results with some advanced ranking methods and observe that the accuracy of our method for all users and cold-start users improves. Our method can also create more items for cold-start users in the list of recommended items.

Fraud, misidentification, and adulteration of food, whether unintentional or purposeful, are a worldwide and growing concern. Aquaculture and fisheries are recognized as one of the sectors most vulnerable to food fraud. Besides, a series of risks related to health and distrust between consumer and popular market makes this sector develop an effective solution for fraud control. Species identification is an essential aspect to expose commercial fraud. Convolutional neural networks (CNNs) are one of the most powerful tools for image recognition and classification tasks. Thus, the objective of this study is to propose a model of recognition of fish species based on CNNs. After the implementation and comparison of the results of the CNNs, it was found that the Xception architecture achieved better performance with 86% accuracy. It was also possible to build a web application mockup. The proposal is easily applied in other aquaculture areas such as the species recognition of lobsters, shrimp, among other seafood.

Despite a recent surge of research interest in privacy and transferrable deep learning, optimizing the tradeoff between privacy requirements and performance of machine learning models remains a challenge. This motivates the development of an approach that optimizes both privacy-preservation mechanism and learning of the deep models for achieving a robust performance. This paper considers the problem of semi-supervised transfer and multi-task learning under differential privacy framework. An alternative conception of deep autoencoder, referred to as Conditionally Deep Membership-Mapping Autoencoder (CDMMA), is considered for transferrable deep learning. Under practice-oriented settings, an analytical solution for the learning of CDMMA can be derived by means of variational optimization. The paper proposes a transfer and multi-task learning approach that combines CDMMA with a tailored noise adding mechanism to transfer knowledge from source to target domain in a privacy-preserving manner.

Correct classification of neuromuscular disorders is essential to provide accurate diagnosis. Presently, gene microarray technology is a widely accepted technology to monitor the expression level of a large number of genes simultaneously. The gene microarray data are a high dimensional data, which usually contains small samples having a large number of genes. Therefore, dimension reduction is a crucial task for correct classification of diseases. Dimension reduction eliminates those genes which are less expressive and enhances the efficiency of the classification model. In the present paper, we developed a novel hybrid dimension reduction method and a deep learning-based classification model for neuromuscular disorders. The hybrid dimension reduction method is deployed in three phase: in the first phase, the expressive genes are selected using F test method, and the mutual information method and the best one among them are selected for further processing. In second phase, the gene selected by the best model is further transformed to low dimension by PCA. In third phase, the deep learning-based classification model is deployed. For experimentation, two diseased and multi-diseased micro array data sets, which is publicly available, is used. The best accuracy by 50-100-50-25-13 deep learning architecture with hybrid dimension reduction, where 100 genes select by F test and PCA with 50 principal components is 89% for NMD data set. The best accuracy by 50-100-2 deep learning architecture with hybrid dimension reduction, where 100 genes select by F test and PCA with 50 principal components is 97% for FSHD data set. The proposed hybrid method gives better classification accuracy result and reduces the search space and time complexity as well for both two diseased and multi-diseased micro array data sets.

Convolutional Neural Networks are machine learning models that have proven abilities in many variants of tasks. This powerful machine learning model sometimes suffers from overfitting. This paper proposes a method based on Reinforcement Learning for addressing this problem. In this research, the parameters of a target layer in the Convolutional Neural Network take as a state for the Agent of the Reinforcement Learning section. Then the Agent gives some actions as forming parameters of a hyperbolic secant function. This function’s form is changed gradually and implicitly by the proposed method. The inputs of the function are the weights of the layer, and its outputs multiply by the same weights to updating them. In this study, the proposed method is inspired by the Deep Deterministic Policy Gradient model because the actions of the Agent are into a continuous domain. To show the proposed method’s effectiveness, the classification task is considered using Convolutional Neural Networks. In this study, 7 datasets have been used for evaluating the model; MNIST, Extended MNIST, small-notMNIST, Fashion-MNIST, sign language MNIST, CIFAR-10, and CIFAR-100.