Neural Network Based Prediction Model Research Articles

Design and Modeling of Stock Market Forecasting Using Hybrid Optimization Techniques

In this paper, an artificial neural network-based stock market prediction model was developed. Today, a lot of individuals are making predictions about the direction of the bond, currency, equity, and stock markets. Forecasting fluctuations in stock market values is quite difficult for businesspeople and industries. Forecasting future value changes on the stock markets is exceedingly difficult since there are so many different economic, political, and psychological factors at play. Stock market forecasting is also a difficult endeavour since it depends on so many various known and unknown variables. There are several ways used to try to anticipate the share price, including technical analysis, fundamental analysis, time series analysis, and statistical analysis; however, none of these approaches has been shown to be a consistently reliable prediction tool. We built three alternative Adaptive Neuro-Fuzzy Inference System (ANFIS) models to compare the outcomes. The average of the tuned models is used to create an ensemble model. Although comparable applications have been attempted in the literature, the data set is extremely difficult to work with because it only contains sharp peaks and falls with no seasonality. In this study, fuzzy c-means clustering, subtractive clustering, and grid partitioning are all used. The experiments we ran were designed to assess the effectiveness of various construction techniques used to our ANFIS models. When evaluating the outcomes, the metrics of R-squared and mean standard error are mostly taken into consideration. In the experiments, R-squared values of over.90 are attained.

BP fuzzy neural network PID based constant tension control of traction winch

As part of the ocean winch, the traction winch mainly plays the role of attenuating the load tension. It generally uses simple speed control. For the ship-based collection and release system, complex sea conditions make the cable tension constantly changes. The traditional control method will make the cable in the traction winch too slack or too tight. Meanwhile, it can result in the wrong rope, bite rope and slippage, and other phenomena. Eventually, it is leading to cable damage or breakage and equipment loss. Moreover, with the introduction of expensive photoelectric composite cables, there is a greater need for a good control algorithm to extend the life of the cable and ensure the safety of retrieval and release. Therefore, this paper focuses on the constant tension control of the traction winch cable. In this paper, the NARX neural network-based cable tension compensation value prediction model is established as the feedforward part based on historical ship heave displacement data and cable tension compensation data as training data. On this basis, the simulation model of cable constant tension winch based on common PID control, fuzzy PID control, and BP fuzzy neural network PID control are established. By comparing the tension of the cable with the above controller, the results show that the tension control of the cable based on the NARX neural network predicted tension compensation value plus BP fuzzy neural network PID control has high accuracy, fast dynamic response, smooth system, and no overshoot.

Deep learning-based fully automated intelligent and precise diagnosis for melanocytic lesions

Melanocytic lesions occur on the surface of the skin, in which the malignant type is melanoma with a high fatality rate, seriously endangering human health. The histopathological analysis is the gold standard for diagnosis of melanocytic lesions. In this study, a fully automated intelligent diagnosis method based on deep learning was proposed to classify the pathological whole slide images (WSI) of melanocytic lesions. Firstly, the color normalization based on CycleGAN neural network was performed on multi-center pathological WSI; Secondly, ResNet-152 neural network-based deep convolutional network prediction model was built using 745 WSI; Then, a decision fusion model was cascaded, which calculates the average prediction probability of each WSI; Finally, the diagnostic performance of the proposed method was verified by internal and external test sets containing 182 and 54 WSI, respectively. Experimental results showed that the overall diagnostic accuracy of the proposed method reached 94.12% in the internal test set and exceeded 90% in the external test set. Furthermore, the color normalization method adopted was superior to the traditional color statistics-based and staining separation-based methods in terms of structure preservation and artifact suppression. The results demonstrate that the proposed method can achieve high precision and strong robustness in pathological WSI classification of melanocytic lesions, which has the potential in promoting the clinical application of computer-aided pathological diagnosis.

Continual learning-based trajectory prediction with memory augmented networks

Forecasting pedestrian trajectories is widely used in mobile agents such as self-driving vehicles and social robots. Deep neural network-based trajectory prediction models precisely predict pedestrian trajectories after training. However, the prediction models fail to avoid catastrophic forgetting when the data distribution shifts during continual learning, making it incredible to deploy the models on agents in real environments. A continual trajectory prediction method with memory augmented networks, CLTP-MAN, is proposed to handle the catastrophic forgetting issue by introducing a memory augmented network with sparse experience replay. CLTP-MAN comprises an external memory module, a memory extraction module, and a trajectory prediction module. The external memory module contains prior knowledge useful for trajectory prediction. The memory extraction module can read or write the key–value memories with a trainable controller. Last, the trajectory prediction module performs long-trajectory prediction by introducing a multi-hop attention mechanism to extract pivotal information from the external memory. Meanwhile, the catastrophic forgetting issue is handled through sparse experience replay. The two benchmarking datasets ETH/UCY and SDD are reintegrated according to the needs for continual learning to conduct quantitative and qualitative evaluations. The results verify that benefiting from external memory and the multi-hop attention mechanism, CLTP-MAN has better generalization than several mainstream methods. Sparse experience replay effectively reduces catastrophic forgetting, leading to reliable deployments on mobile agents.

Explicit Neural Network-Based Models for Bubble Point Pressure and Formation Volume Factor Prediction

Explicit Neural Network-Based Models for Bubble Point Pressure and Formation Volume Factor Prediction

Artificial neural networks approach for prediction of CIELab values for yarn after dyeing and finishing process

In textile products, color plays an influential role in changing fashion trends. The main challenge in dyeing processes is the achievement of the desired color output (CIELab value) through various recipe matching algorithms and programs. A color matching approach in this context uses the type of dyestuff, the color of pre-treated fabric, and the owf% value of dyestuff as input sets, and the recipe for dyeing and CIELab values as outputs. The fiber, yarn, and finishing parameters that could influence dyeing processes, however, are excluded from these approaches (based on Kubelka-Munk theory). This study aims to develop a model that can predict the CIELab values (L*, a*, and b*) of dyed yarns after package dyeing. In order to predict CIELab values, artificial neural network-based prediction models have been developed using the properties of cotton fiber used in yarn production (HVI-AFIS etc.), yarn spinning, dyeing and finishing process parameters as input sets. A study has also been carried out on how input sets are to be related to the optimization of a prediction model. In order to account for the variability of the input sets, two different prediction models have been developed. Both models yield low prediction errors for three outputs: L*, a*, and b*. Based on the predictions by the models, the difference between the actual values and the predicted values is found to be within acceptable tolerances for L*, a* and b*, and a large part of the 498 predicted test data have an absolute error deviation of less than 0.5.

Prediction of the Effect of Nutrients on Plant Parameters of Rice by Artificial Neural Network

Rice holds key importance in food and nutritional security across the globe. Nutrient management involving rice has been a matter of interest for a long time owing to the unique production environment of rice. In this research, an artificial neural network-based prediction model was developed to understand the role of individual nutrients (N, P, K, Zn, and S) on different plant parameters (plant height, tiller number, dry matter production, leaf area index, grain yield, and straw yield) of rice. A feed-forward neural network with back-propagation training was developed using the neural network (nnet) toolbox available in Matlab. For the training of the model, data obtained from two consecutive crop seasons over two years (a total of four crops of rice) were used. Nutrients interact with each other, and the resulting effect is an outcome of such interaction; hence, understanding the role of individual nutrients under field conditions becomes difficult. In the present study, an attempt was made to understand the role of individual nutrients in achieving crop growth and yield using an artificial neural network-based prediction model. The model predicts that growth parameters such as plant height, tiller number, and leaf area index often achieve their maximum performance at below the maximum applied dose, while the maximum yield in most cases is achieved at 100% N, P, K, Zn, and S dose. In addition, the present study attempted to understand the impact of individual nutrients on both plant growth and yield in order to optimize nutrient recommendation and nutrient management, thereby minimizing environmental pollution and wastage of nutrients.

A Study on Cryptocurrency Log-Return Price Prediction Using Multivariate Time-Series Model

Cryptocurrencies are highly volatile investment assets and are difficult to predict. In this study, various cryptocurrency data are used as features to predict the log-return price of major cryptocurrencies. The original contribution of this study is the selection of the most influential major features for each cryptocurrency using the volatility features of cryptocurrency, derived from the autoregressive conditional heteroskedasticity (ARCH) and generalized autoregressive conditional heteroskedasticity (GARCH) models, along with the closing price of the cryptocurrency. In addition, we sought to predict the log-return price of cryptocurrencies by implementing various types of time-series model. Based on the selected major features, the log-return price of cryptocurrency was predicted through the autoregressive integrated moving average (ARIMA) time-series prediction model and the artificial neural network-based time-series prediction model. As a result of log-return price prediction, the neural-network-based time-series prediction models showed superior predictive power compared to the traditional time-series prediction model.

Research on Quantitative Trading Investment Strategy Based on LSTM and Dynamic Programming

As financial markets matured, more standardized and quantitative research emerged. In this paper, we model the historical price returns of U.S. gold and BTC over a five-year trading period from September 11, 2016, to September 10, 2021, and estimate the total investment returns as of September 10, 2021. First, we build a variety of price-prediction models to fit the image to the data to get the most accurate image of the actual price trend. It was found that the LSTM neural network-based multi-interval segmentation price prediction model, which has the best prediction effect by comparison with the actual value. Therefore, it was used as the core prediction model for this paper. Then, we use the planning model to find the optimal investment plan by using the difference between the price change of gold and bitcoin trading in the next three days as the decision variable, the highest Sharpe ratio on the third day as the objective function. At the same time, the variance of the forecast value in the next three days is used as the risk characterization, and different weights are assigned to the risk and return as the objective function to characterize the investment strategy under different investment personalities. In order to accelerate the convergence of the planning model, we added a particle swarm algorithm for optimization. In the end, we obtained the results for the specific daily investment scenarios for aggressive investors, intermediate investors, and steady investors.

Short Term Load Forecasting of Residential and Commercial Consumers of Karnataka Electricity Board using CFNN

Electricity use and its access are correlated in the economic development of any country. Economically, electricity cannot be stored, and for stability of an electrical network a balance between generation and consumption is necessary. Electricity demand depends on various factors like temperature, everyday activities, time of day, days of the week days/Holidays. These parameters have led to price volatility and huge spikes in electricity prices. The research work proposes a short term Load prediction Model for LT2 (residential consumers), LT3 (Commercial Consumers) of Karnataka State Electricity Board using Cascaded Feed Forward Neural Network (CFNN). MATLAB software is utilized to design and test the forecasting model for predicting the power consumption. Furthermore, a shallow feed forward neural network-based prediction model is constructed and evaluated for performance comparison. The Performance metrics include Mean Absolute Percentage Error (MAPE) and Mean Squared Error (MSE). The suggested STLF CFNN prediction model outperformed shallow feed forward networks on both performance metrics with prediction errors of less than 1%.

Prediction of the stability number of conventional rubble-mound breakwaters using machine learning algorithms

An important issue in designing the structures of rubble-mound breakwaters, is to estimate the stability number of its armor block. Most of the traditional stability analysis methods are not compatible enough to handle the obscurities, indistintness and uncertainties of this field. The relations between stability number, damage level and other stability variables can better be modeled using the advanced techniques of machine learning (ML) algorithms. In this prospect, three new ML models consisting of two ensemble learning models; Random Forest, Gradient Boosting and one fully connected deep artificial neural network based prediction model have been presented in this study. Using the ensemble learning models a detailed feature analysis has been introduced here, to understand the feature importances of stability variables on the stability number. To the best of the author’s knowledge, these have never been used in this field of stability analysis of rubble-mound breakwaters. Outperforming all of the conventional methods, the proposed study has delivered the highest level of accuracy as 99%, in the prediction of the stability number. Also, the proposed ML models are found to perform better, in dealing with the complex non-linearities related to this field. The feature analysis gives a meaningful insight into the dataset. Therefore, this study can be a useful alternative approach for the designers of the rubble-mound breakwaters.

Deep Learning and Neural Network-Based Wind Speed Prediction Model

This paper aims to develop a wind speed prediction model by utilizing deep learning and neural networks. The analysis of weather data using a neural network architecture has been completed. The Long Short-Term Memory (LSTM) architecture is a type of artificial Recurrent Neural Network (RNN) used in deep learning is the first method plots the predicting Wind Speed based on the dataset and predicts the future spread. A dataset from a real-time weather station is used in the implementation model. The dataset consists of information from the weather station implements of the recurrent neural network model that plots the past spread and predicts the future stretch of the weather. The performance of the recurrent neural network model is presented and compared with Adaline neural network, Autoregressive Neural Network (NAR), and Group Method of Data Handling (GMDH). The NAR used three hidden layers. The performance of the model is analyzed by presenting the Wind Speeds of Erbil city. The dataset consists of the Wind Speed of (1992-2020) years, and each year consist of twelve months (from January to December).

Natural Frequency Degradation Prediction for Offshore Wind Turbine Structures

The scale of offshore wind turbines (OWTs) has increased in order to enhance their energy generation. However, strong aero/hydrodynamic loads can degrade the dynamic characteristics of OWTs because they are installed on soft seabeds. This degradation can shorten the structural life of the system; repetitive loads lead to seabed softening, reducing the natural frequency of the structure close to the excitation frequency. Most of the previous studies on degradation trained prediction algorithms with actual sensor signals. However, there are no actual sensor data on the dynamic response of OWTs over their lifespan (approximately 20 years). In order to address this data issue, this study proposes a new prediction platform combining a dynamic OWT model and a neural network-based degradation prediction model. Specifically, a virtual dynamic response was generated using a three-dimensional OWT and a seabed finite element model. Then, the LSTM model was trained to predict the natural frequency degradation using the dynamic response as the model input. The results show that the developed model can accurately predict natural frequencies over the next several years using past and present accelerations and strains. In practice, this LSTM model could be used to predict future natural frequencies using the dynamic response of the structure, which can be measured using actual sensors (accelerometers and strain gauges).

Artificial neural network-based roughness prediction models for gravel roads considering land use

Artificial neural network-based roughness prediction models for gravel roads considering land use

Influence of mass-awareness campaign on community behavior pattern changes for safe drinking water availability in a groundwater arsenic-affected area of South Asia

Environmental pollution of drinking water and it's mitigation policies spose one of the most significant global sustainability challenges. The remediation of pollutants from drinking water depends on technological efficacy, whereas its implementation on the target population requires social acceptance besides economic resources. In this context, mass awareness of the benefits of community safe water supply scheme and people's willingness to use the treated water may improvise the effective implementation of the project. The present research demonstrated the effectiveness of a field-scale mass awareness experiment on the management of arsenic (As) pollution through the implementation of piped community water supply project, in areas with profuse As contaminated tubewell-based water use. Extensive baseline and end-line surveys were conducted to assess the public awareness of the As issues, identification of the captive population, which delineates the impact of the campaign program and the associated change in behavior of the studied population. Lorenz curve was employed to estimate the awareness of arsenic contamination in different population clusters and it was observed that a cumulative awareness share of 96.9% was achieved in 97.5% of households. Hotspot analysis was conducted to identify the spatial distribution of As incidences and unresponsive captive population to awareness program. A neural network-based prediction model was developed, integrating the baseline and end-line survey data to assess the impact of the awareness program on public behavioral change on shifting to community piped water supply sources from contaminated groundwater. The high value of R2 (0.986) and F statistics (3721.6) signify the effective application of the model

A Grey BP Neural Network-Based Model for Prediction of Court Decision Service Rate.

The judgment service rate is an important index to reflect the fairness of the judgment of legal cases in a certain area, which is of great significance to verify the accuracy of a court judgment. In this paper, a grey neural network model combining grey system theory and BP neural network algorithm is proposed to predict the index. Analyze the judgment service rate of the court judgment system, and build a prediction system based on the completion rate, completion rate, plaintiff satisfaction, defendant satisfaction, litigation time, property preservation cycle, document delivery time, implementation information disclosure rate, and other key indicators. Through example analysis, it is proved that the combined model of the grey prediction model and BP neural network has a small error and good simulation effect on the prediction of court decision-making service rate, which can better promote the development of court and society.

Neural Network-Based Prediction Model for the Stability of Unlined Elliptical Tunnels in Cohesive-Frictional Soils

The scheme for accurate and reliable predictions of tunnel stability based on an artificial aeural network (ANN) is presented in this study. Plastic solutions of the stability of unlined elliptical tunnels in sands are first derived by using numerical upper-bound (UB) and lower-bound (LB) finite element limit analysis (FELA). These numerical solutions are later used as the training dataset for an ANN model. Note that there are four input dimensionless parameters, including the dimensionless overburden factor γD/c′, the cover–depth ratio C/D, the width–depth ratio B/D, and the soil friction angle ϕ. The impacts of these input dimensionless parameters on the stability factor σs/c′ of the stability of shallow elliptical tunnels in sands are comprehensively examined. Some failure mechanisms are carried out to demonstrate the effects of all input parameters. The solutions will reliably and accurately provide a safety assessment of shallow elliptical tunnels.

ANN-Based Inverse Goal-Oriented Design Method for Targeted Final Properties of Materials

Designing materials for targeted materials properties is the key to tackle the demands for personalized consumer products. The deficiency in the existing linear and nonlinear correlation methods attributed to simplifying assumptions and idealizations, nondeterministic simulations, and limited experimental data due to heavy computational time and cost, necessitates a design method that provides sufficient confidence to designers in decision making. To address this requirement, we propose, in this paper, an inverse goal-oriented materials design method supported by the design space exploration framework (DSEF). Keeping in view the accuracy and precision in the prediction confidence of machine learning-based methods, we developed an Artificial Neural Network based prediction model that supports DSEF. The proposed method for materials design can help designers to (1) explore PSPP spaces starting from end property requirements, (2) adjust the errors being propagated in the PSPP chain as well as in the predictions made by the model, and (3) timely adjust model parameters of the prediction model for accurate predictions. The efficacy of the method is illustrated for the hot stamping process to produce structural components from ultrahigh-strength steels (UHSS). The proposed method and prediction model are generic and applicable to any sequential manufacturing process to realize an end product.

Reliability Characterization of Solder Joints in Electronic Systems Through a Neural Network Aided Approach

Rapid and precise reliability evaluation of electronic circuits plays a key role in the design stage of the electronic systems. The task becomes even more difficult when several major parameters contribute into the reliability evaluation. This paper proposes a neural network aided approach as a prediction tool for estimating the useful lifetime of the ball shaped solder joint as the most resistless part under accidental drops in the electronic devices. Several contributory factors including ball grid array (BGA) chip location, printed circuit board (PCB) thickness, solder alloy composition and solder ball volume are considered in our proposed rapid prediction model and their effects are investigated. 480 finite element simulations as well as 20 experimental tests are performed to obtain an enriched database for our neural network based prediction model. The accuracy of the proposed model is calculated as 97.55% in comparison with the finite element and the experimental results. Ability of considering multi contributory factors in the reliability evaluation of the BGA chip makes our proposed approach be a suitable candidate in design for reliability of the electronic systems.

An in-Depth Analysis of the Software Features’ Impact on the Performance of Deep Learning-Based Software Defect Predictors

<i>Software Defects Prediction</i> represents an essential activity during software development that contributes to continuously improving <i>software quality</i> and software maintenance and evolution by detecting defect-prone modules in new versions of a software system. In this paper, we are conducting an in-depth analysis on the software features&#x2019; impact on the performance of deep learning-based software defect predictors. We further extend a large-scale feature set proposed in the literature for detecting defect-proneness, by adding conceptual software features that capture the semantics of the source code, including comments. The conceptual features are automatically engineered using Doc2Vec, an artificial neural network based prediction model. A broad evaluation performed on the Calcite software system highlights a statistically significant improvement obtained by applying deep learning-based classifiers for detecting software defects when using conceptual features extracted from the source code for characterizing the software entities.