改进的多层感知机在客户流失预测中的应用

针对在传统的客户流失预测数据预处理中，使用one-hot编码处理离散属性导致数据维度增加及数据过于稀疏的问题，提出了两种基于多层感知机的改进后的客户流失预测模型。其主要思想是分别使用堆叠自编码器和实体嵌入两种方法对多层感知机进行改进，通过将离散属性的高维编码数据向低维空间映射，有效地减少了one-hot编码产生的稀疏数据，增加了离散属性值之间的关联度。在对两份公开的数据集进行交叉验证后的实验结果表明，改进后的模型既有效地提高了预测的准确度，又维持了传统多层感知机模型在并行化计算方面的优势。     

基于多层感知器的气液两相流流型识别方法

通过对电阻层析成像数据采集原理和深度学习网络的研究,提出了一种基于阵列电阻值和多层感知器深度学习网络相结合的流型识别方法。利用电阻层析成像系统中的16个电极传感器来获取流型样本数据,并构建出流型识别数据库,然后对多层感知器深度学习网络进行训练,获得可以识别不同流型的网络。实验结果表明,采用阵列电阻值结合多层感知器网络对流型进行学习和识别的方法,流型识别准确率可以达到95%,解决了流型图像生成过程与数据特征预选过程中流型特征损失的问题,流型识别性能得到了提高。     

DEM: Deep Entity Matching Across Heterogeneous Information Networks

Heterogeneous information networks, which consist of multi-typed vertices representing objects and multi-typed edges representing relations between objects, are ubiquitous in the real world. In this paper, we study the problem of entity matching for heterogeneous information networks based on distributed network embedding and multi-layer perceptron with a highway network, and we propose a new method named DEM short for Deep Entity Matching. In contrast to the traditional entity matching methods, DEM utilizes the multi-layer perceptron with a highway network to explore the hidden relations to improve the performance of matching. Importantly, we incorporate DEM with the network embedding methodology, enabling highly efficient computing in a vectorized manner. DEM’s generic modeling of both the network structure and the entity attributes enables it to model various heterogeneous information networks flexibly. To illustrate its functionality, we apply the DEM algorithm to two real-world entity matching applications: user linkage under the social network analysis scenario that predicts the same or matched users in different social platforms and record linkage that predicts the same or matched records in different citation networks. Extensive experiments on real-world datasets demonstrate DEM’s effectiveness and rationality.

     

Artificial Neural Networks for Prediction of COVID-19 in Saudi Arabia

In this study, we have proposed an artificial neural network (ANN) model to estimate and forecast the number of confirmed and recovered cases of COVID-19 in the upcoming days until September 17, 2020. The proposed model is based on the existing data (training data) published in the Saudi Arabia Coronavirus disease (COVID-19) situation—Demographics. The Prey-Predator algorithm is employed for the training. Multilayer perceptron neural network (MLPNN) is used in this study. To improve the performance of MLPNN, we determined the parameters of MLPNN using the prey-predator algorithm (PPA). The proposed model is called the MLPNN–PPA. The performance of the proposed model has been analyzed by the root mean squared error (RMSE) function, and correlation coefficient (R). Furthermore, we tested the proposed model using other existing data recorded in Saudi Arabia (testing data). It is demonstrated that the MLPNN-PPA model has the highest performance in predicting the number of infected and recovering in Saudi Arabia. The results reveal that the number of infected persons will increase in the coming days and become a minimum of 9789. The number of recoveries will be 2000 to 4000 per day.     

Learning machines: Rationale and application in ground-level ozone prediction

Multilayer perceptron (MLP) and support vector machine (SVM), two popular learning machines, are increasingly being used as alternatives to classical statistical models for ground-level ozone prediction. However, employing learning machines without sufficient awareness about their limitations can lead to unsatisfactory results in modeling the ozone evolving mechanism, especially during ozone formation episodes. With the spirit of literature review and justification, this paper discusses, with respect to the concerning of ozone prediction, the recently developed algorithms/technologies for treating the most prominent model-performance-degradation limitations. MLP has the “black-box” property, i.e., it hardly provides physical explanation for the trained model, overfitting and local minima problems, and SVM has parameters identification and class imbalance problems. This commentary article aims to stress that the underlying philosophy of using learning machines is by no means as trivial as simply fitting models to the data because it causes difficulties, controversies or unresolved problems. This article also aims to serve as a reference point for further technical readings for experts in relevant fields.     

Parkinson’s Disease Detection Using Biogeography-Based Optimization

In recent years, Parkinson's Disease (PD) as a progressive syndrome of the nervous system has become highly prevalent worldwide. In this study, a novel hybrid technique established by integrating a Multi-layer Perceptron Neural Network (MLP) with the Biogeography-based Optimization (BBO) to classify PD based on a series of biomedical voice measurements. BBO is employed to determine the optimal MLP parameters and boost prediction accuracy. The inputs comprised of 22 biomedical voice measurements. The proposed approach detects two PD statuses: 0-disease status and 1- good control status. The performance of proposed methods compared with PSO, GA, ACO and ES method. The outcomes affirm that the MLP-BBO model exhibits higher precision and suitability for PD detection. The proposed diagnosis system as a type of speech algorithm detects early Parkinson’s symptoms, and consequently, it served as a promising new robust tool with excellent PD diagnosis performance.     

Estimating of the Dry Unit Weight of Compacted Soils Using General Linear Model and Multi-layer Perceptron Neural Networks

Compaction of earth fill is a very important stage of construction projects. Degree of compaction is defined by relative compaction. The relative compaction of a compacted earth fill is calculated by dividing the dry unit weight obtained from in situ tests by-into the maximum dry unit weight obtained from laboratory compaction tests. This rate represents compaction quality in the field. Numerous test methods such as sand cone, rubber balloon, nuclear measurements, etc., are available to determine the maximum dry unit weight of soils in the field. It is well known that these methods have disadvantages as well as advantages. This study focused on estimation of dry unit weight of soils depending on water contents and P-wave velocities of compacted soils. The multi-layer perceptron (MLP) neural networks and general linear model (GLM) were used in this study to estimate the dry unit weight of different types of soils. Results of the MLP neural networks were compared with the GLM results. Based on the comparisons, it is found that the MLP generally gives better dry unit weight estimates than the GLM technique. The laboratory experiments and modeling studies showed that a new method for compaction control can be developed depending on P-wave velocity to estimate of the dry unit weight of compacted soils.     

An exergetically-sustainable operational condition of a photo-biohydrogen production system optimized using conventional and innovative fuzzy techniques

The aim of the present study was to perform an exergy-based multi-objective fuzzy optimization of a continuous photobioreactor applied for biohydrogen production from syngas via the water-gas shift reaction by Rhodospirillum rubrum. For this purpose, the conventional and innovative fuzzy optimization techniques coupled with multilayer perceptron (MLP) neural model to optimize the main exergetic performance parameters of the photobioreactor. The MLP neural model was applied to correlate three dependent variables (rational and process exergy efficiencies and normalized exergy destruction) with two independent variables (syngas flow rate and agitation speed). The developed MLP model was then interfaced with three different multi-objective fuzzy optimization systems with independent, interdependent, and locally modified interdependent objectives. The optimization process was aimed at maximizing the rational exergy and process efficiencies, while minimizing the normalized exergy destruction, simultaneously. Generally, the innovative locally modified interdependent objectives fuzzy system showed a better optimization capabilities compared with the other two fuzzy systems. Accordingly, the optimal syngas photo-fermentation for biohydrogen production in the continuous bioreactor corresponded to the agitation speed of 383.34 rpm and syngas flow rate of 13.35 mL/min in order to achieve the normalized exergy destruction of 1.56, rational exergy efficiency of 85.65%, and process exergy efficiency of 21.66%.     

Modelling photovoltaic modules with neural networks using angle of incidence and clearness index

A neural network for modelling photovoltaic modules using angle of incidence and clearness index is proposed. Engineers require methods to estimate the output of a photovoltaic plant depending on meteorological conditions. Therefore, models for the grid inverter and the generator must be provided, and their outputs must be combined. The connection between both models is related to the maximum power point of the generator and how it is tracked by the inverter. That maximum power point under specific conditions of irradiance and module temperature is determined by the I–V curve of the module, which must be simulated under those conditions. Algebraic procedures were used to simulate the I–V curve. Recently, neural networks have been used for the same purpose. Previous methods only take into account the irradiance and the module temperature. The model proposed is based on neural networks, and it uses not only the irradiance and the module temperature but also the angle of incidence and the instantaneous clearness index as additional inputs. The normalised clearness replaces the standard clearness index because it allows the removal of the hourly trend found in this last index. This new model improves the results obtained with previous ones as it can distinguish amongst samples with the same solar irradiance and temperature values but with different angle of incidence and instantaneous clearness index. Results show that this model could be used to improve the accuracy of the tools used to forecast the output of photovoltaic plants. Copyright © 2014 John Wiley & Sons, Ltd.