Mean Absolute Error Values Research Articles

Novel Explicit Models for Assessing the Frictional Resistance of Pipe Piles Subjected to Seismic Effects

This paper introduces novel explicit models to predict the frictional resistance of open and closed-ended pipe piles subjected to seismic loading. This research employs genetic programming (GP) and multiobjective genetic algorithm-based evolutionary polynomial regression (EPR-MOGA) to develop closed-form expressions for estimating pile frictional resistance, utilizing widely used input parameters for enhanced practicality and applicability in engineering practice. The proposed models are developed using only three input variables: the corrected standard penetration test (SPT) blow count (N1)60, the pile slenderness ratio (L/D), and the peak ground acceleration (PGA). This deliberate reduction in input complexity significantly enhances the models' applicability across a wide range of geotechnical scenarios and industries. The accuracy of the developed models was assessed via the coefficient of determination (R2), root mean squared error (RMSE), and mean absolute error (MAE). In the case of the GP model, the evaluation metrics for the testing set for open-ended piles (R2, RMSE, and MAE values) are 0.89, 0.43, and 0.35, respectively, whereas the corresponding values for closed-ended piles are 0.93, 0.38, and 0.3, respectively. On the other hand, the EPR-MOGA approach achieves similarly encouraging results, with performance metrics of 0.92, 0.37, and 0.29 for open-ended piles and 0.91, 0.39, and 0.30 for closed-ended piles.

Prediction and analysis of relative error in electric vehicle charging stations based on an improved ConvFormer model

In order to address the issue of metering inaccuracies in charging stations that directly affect the development of electric vehicles, a prediction method for the relative error of charging stations based on the ConvFormer model is proposed. The model combines Convolutional Neural Networks (CNN) with Transformer models in parallel, significantly improving the prediction accuracy. First, charging station data is preprocessed using forward interpolation and normalization methods, and the dataset is transformed into a dataset of input relative errors. Then, a neural network with an improved unidirectional convolutional and attention combination for time-series forecasting is constructed, and common regression performance evaluation metrics, MAE (Mean Absolute Error) and MSE (Mean Squared Error), are selected for evaluation. Finally, based on seven days of charging station data, the relative error of charging stations for the next 24 h is predicted, and compared to traditional Transformer and LSTM (Long Short-Term Memory) time-series models. The results show that the improved model yields the lowest values for both MAE and MSE, with a 47.30 % reduction in MAE compared to the Transformer model and a 38.06 % reduction compared to LSTM, and a 66.94 % reduction in MSE compared to the Transformer model and approximately 62.32 % reduction compared to LSTM.

Evaluation of the collapse susceptibility of loess using machine learning

Evaluating collapse susceptibility of loess is essential for the construction of transportation lines in loess regions as it provides guidance for ground treatment. For the existing methods, a large number of boreholes need to be drilled along transportation lines to collect intact samples for laboratory tests, which make them very time and cost-consuming. In this study, loess’s collapse susceptibility is evaluated using Multi Expression Programming (MEP) and Back-Propagation Neural Network (BPNN). According to analysis of wetting-induced loess collapse, the gravimetric water content at the initial state (w0), net vertical stress (σ-ua), void ratio at the initial state (e0), void ratio at the liquid limit state (eL), and plastic index (Ip) are chosen as input variables. A comprehensive database incorporating 200 oedometrically soaking tests is established to train and test the two algorithms. The collapse potentials of loess are well predicted using MEP and BPNN, as demonstrated by high values of coefficient of determination (R2>0.88) and small values of mean absolute error (MAE<0.008) and root mean squared error (RMSE<0.012). Following ASTM D5333-03 [5], the degree of loess collapse is classified with accuracies of 98 % and 90 % for MEP and BNPP respectively. Furthermore, sensitivity analysis shows the contribution of each variable to the prediction of collapse potential follows the order of e0>eL>Ip>σ-ua>w0. The machine learning is expected to assist the code of practice ASTM D5333-03 [5] in achieving an efficient site-investigation of collapsible loess for the construction of transportation lines.

Remaining useful life prediction for a cracked rotor system via moving feature fusion based deep learning approach

This study proposes a novel approach for predicting the remaining useful life (RUL) of a cracked rotor system by using a moving fusion gated recurrent unit (MFGRU) model. Firstly, 12 multi-domain features were extracted from the raw collected vibration signals, which were then fused and distributed adaptively using bidirectional exponential moving average (EMA) and multi-head attention (MHA). Then a bidirectional gated recurrent unit network combined with moving feature fusion method was proposed to capture the long-term dependence relationship in the monitoring data of the cracked rotor for RUL prediction. Finally, the model’s performance was validated through accelerated life experiments, with the measured values of root mean square error (RMSE) and mean absolute error (MAE) below 3.17 and 2.60, respectively. This study offers a dynamic RUL prediction method with certain significance and valuable reference for designing deep learning models.

HYBRID MODEL OF SINGULAR SPECTRUM ANALYSIS WITH AUTOREGRESSIVE INTEGRATED MOVING AVERAGE AND FUZZY TIME SERIES FOR INDONESIAN CRUDE PRICE FORECASTING

This study discusses a hybrid model of Singular Spectrum Analysis (SSA) with Autoregressive Integrated Moving Average (ARIMA) and Fuzzy Time Series (FTS) for forecasting the Indonesian Crude Price (ICP). SSA is considered to capture the deterministic component of the data while the ARIMA and FTS are to represent the stochastics one. The data that used in this study are ICP per month from January 2017 to May 2023. The data from January 2017 to December 2022 are used as insample data, while the data from January to May 2023 are used as outsample data. The insample data is firstly modeled by SSA and the residuals are then modeled by ARIMA, referred to as the hybrid SSA-ARIMA. By the same procedure, the hybrid SSA-FTS model is also constructed to the insample data. Based on the experiment, the hybrid SSA-ARIMA produces Mean Absolute Percentage Error values 8.08% for an insample and 7.10% for an outsample data. These values are less than those obtained by hybrid SSA-FTS. Therefore, the hybrid SSA-ARIMA is recommended for forecasting the monthly ICP.

Prediction of the level of crime cases using multiple linear regression in the city of Pontianak

This study aims to develop a predictive model for the crime rate in the Police Resort Area of Kota (POLRESTA) Pontianak using the Multiple Linear Regression method based on secondary data obtained from the Criminal Investigation Unit of POLRESTA Pontianak. The utilization of descriptive statistical techniques and data visualization aids in identifying relevant features that enrich the information within the model. The evaluation results indicate that this model performs well in both modeling and predicting crime rates in Kota Pontianak. Despite the variations in error rates between training and testing data, the model still demonstrates its proficiency in predicting known data. The testing results also reveal that the Mean Absolute Percentage Error (MAPE) values for each crime category exhibit variations in the testing dataset, with MAPE for "Berat" increasing to 12.91%, MAPE for "Sedang" increasing to 30.11%, and MAPE for "Ringan" increasing to 26.59%. Consequently, this study concludes that the Multiple Linear Regression method holds potential as an effective tool for decision-making and the development of strategies to combat criminal activities in Kota Pontianak

Findings on celestial pole offsets predictions in the second earth orientation parameters prediction comparison campaign (2nd EOP PCC)

In 2021, the International Earth Rotation and Reference Systems Service (IERS) established a working group tasked with conducting the Second Earth Orientation Parameters Prediction Comparison Campaign (2nd EOP PCC) to assess the current accuracy of EOP forecasts. From September 2021 to December 2022, EOP predictions submitted by participants from various institutes worldwide were systematically collected and evaluated. This article summarizes the campaign's outcomes, concentrating on the forecasts of the dX, dY, and dψ, dε components of celestial pole offsets (CPO). After detailing the campaign participants and the methodologies employed, we conduct an in-depth analysis of the collected forecasts. We examine the discrepancies between observed and predicted CPO values and analyze their statistical characteristics such as mean, standard deviation, and range. To evaluate CPO forecasts, we computed the mean absolute error (MAE) using the IERS EOP 14 C04 solution as the reference dataset. We then compared the results obtained with forecasts provided by the IERS. The main goal of this study was to show the influence of different methods used on predictions accuracy. Depending on the evaluated prediction approach, the MAE values computed for day 10 of forecast were between 0.03 and 0.16 mas for dX, between 0.03 and 0.12 mas for dY, between 0.07 and 0.91 mas for dψ, and between 0.04 and 0.41 mas for dε. For day 30 of prediction, the corresponding MAE values ranged between 0.03 and 0.12 for dX, and between 0.03 and 0.14 mas for dY. This research shows that machine learning algorithms are the most promising approach in CPO forecasting and provide the highest prediction accuracy (0.06 mas for dX and 0.08 mas for dY for day 10 of prediction).Graphical abstract

A Hybrid Deep Learning Model to Estimate the Future Electricity Demand of Sustainable Cities

Rapid population growth, economic growth, and technological developments in recent years have led to a significant increase in electricity consumption. Therefore, the estimation of electrical energy demand is crucial for the planning of electricity generation and consumption in cities. This study proposes a hybrid deep learning model that combines convolutional neural network (CNN) and long short-term memory (LSTM) techniques, both of which are deep learning techniques, to estimate electrical load demand. A hybrid deep learning model and LSTM model were applied to a dataset containing hourly electricity consumption and meteorological information of a city in Türkiye from 2017 to 2021. The results were evaluated using mean absolute percent error (MAPE), root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R2) metrics. The proposed CNN-LSTM hybrid model was compared to the LSTM model, with lower MAPE, MAE, and RMSE values. Furthermore, the CNN-LSTM model exhibited superior prediction performance with an R2 value of 0.8599 compared to the LSTM model with an R2 value of 0.8086. These results demonstrate the effectiveness of the proposed deep learning model in accurately estimating future electrical load demand to plan electricity generation for sustainable cities.

Incorporating Honey Badger Algorithm in Estimating Gamma Distribution With Application to Stock Price Modelling

This study evaluates the performance of various estimation methods in stock price analysis across diverse parameters, focusing on the Honey Badger Algorithm (HBA). The purpose is to determine the most accurate and reliable method for parameter estimation. Methodologically, we analyze data spanning eight years from publicly traded Malaysian property companies, employing financial metrics such as Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). Our findings highlight HBA’s consistent precision in parameter estimation, with values closely aligning with initial parameters across different stock sizes. For example, HBA-Gamma model achieves an MAE of 0.0592 and an RMSE of 0.8458 for 13 stocks, demonstrating its proficiency in capturing stock price distributions in dynamic markets. In contrast, the Artificial Immune System (AIS) provides reasonable estimates but with higher variability. The Regression Method exhibits mixed outcomes, displaying accuracy in some cases but notable variability and reduced precision, especially with larger datasets. The Moment Method, while adequate, shows slightly higher variance compared to both HBA and AIS. Further analysis using Log Likelihood values confirms HBA’s superior fit to the data, consistently surpassing AIS, Regression Method, and Moment Method in likelihood maximization across various stock numbers. Specifically, HBA exhibits lower MAE and RMSE values of 0.1034 and 0.06723, respectively, for 26 stocks, further validating its effectiveness in parameter estimation and stock price prediction. These findings underscore the importance of integrated approaches that account for market nuances rather than relying solely on individual model forecasts. The results affirm HBA’s potential for informed investment decision-making, emphasizing its robust performance and enhanced predictive capabilities compared to alternative methodologies. However, further research is needed to assess the generalizability of these findings to other markets and contexts.

Optimizing Recurrent Neural Networks: A Study on Gradient Normalization of Weights for Enhanced Training Efficiency

Recurrent Neural Networks (RNNs) are classical models for processing sequential data, demonstrating excellent performance in tasks such as natural language processing and time series prediction. However, during the training of RNNs, the issues of vanishing and exploding gradients often arise, significantly impacting the model’s performance and efficiency. In this paper, we investigate why RNNs are more prone to gradient problems compared to other common sequential networks. To address this issue and enhance network performance, we propose a method for gradient normalization of network weights. This method suppresses the occurrence of gradient problems by altering the statistical properties of RNN weights, thereby improving training effectiveness. Additionally, we analyze the impact of weight gradient normalization on the probability-distribution characteristics of model weights and validate the sensitivity of this method to hyperparameters such as learning rate. The experimental results demonstrate that gradient normalization enhances the stability of model training and reduces the frequency of gradient issues. On the Penn Treebank dataset, this method achieves a perplexity level of 110.89, representing an 11.48% improvement over conventional gradient descent methods. For prediction lengths of 24 and 96 on the ETTm1 dataset, Mean Absolute Error (MAE) values of 0.778 and 0.592 are attained, respectively, resulting in 3.00% and 6.77% improvement over conventional gradient descent methods. Moreover, selected subsets of the UCR dataset show an increase in accuracy ranging from 0.4% to 6.0%. The gradient normalization method enhances the ability of RNNs to learn from sequential and causal data, thereby holding significant implications for optimizing the training effectiveness of RNN-based models.

Spectro-ViT: A vision transformer model for GABA-edited MEGA-PRESS reconstruction using spectrograms

This study investigated the use of a Vision Transformer (ViT) for reconstructing GABA-edited Magnetic Resonance Spectroscopy (MRS) data from a reduced number of transients. Transients refer to the samples collected during an MRS acquisition by repeating the experiment to generate a signal of sufficient quality. Specifically, 80 transients were used instead of the typical 320 transients, aiming to reduce scan time. The 80 transients were pre-processed and converted into a spectrogram image representation using the Short-Time Fourier Transform (STFT). A pre-trained ViT, named Spectro-ViT, was fine-tuned and then tested using in-vivo GABA-edited MEGA-PRESS data. Its performance was compared against other pipelines in the literature using quantitative quality metrics and estimated metabolite concentration values, with the typical 320-transient scans serving as the reference for comparison. The Spectro-ViT model exhibited the best overall quality metrics among all other pipelines against which it was compared. The metabolite concentrations from Spectro-ViT's reconstructions for GABA+ achieved the best average R2 value of 0.67 and the best average Mean Absolute Percentage Error (MAPE) value of 9.68%, with no significant statistical differences found compared to the 320-transient reference. The code to reproduce this research is available at https://github.com/MICLab-Unicamp/Spectro-ViT

Utilizing innovative input data and ANN modeling to predict atmospheric gross beta radioactivity in Spain

This study introduces a new methodology aimed at predicting gross β levels in the atmosphere. The methodology incorporates input data consisting of local meteorological and synoptic variables, alongside temporal lags and residence time of air masses, to predict gross β activity concentration in the atmosphere. Weekly measurements conducted between January 2006 and December 2017 at various sampling sites across diverse locations with different climatic and geographical conditions in Spain were utilized. A high-performance Artificial Neural Network (ANN) model was constructed for this purpose. Across all locations, strong linear relationships are evident between predicted and actual values, with correlation coefficients (R) ranging from 0.86 to 0.92. Higher R values indicate a more robust correlation. Additionally, R-squared values, ranging from 0.7320 to 0.8502, further affirm the model’s ability to explain a significant proportion of the variance in gross β activity. Moreover, the relatively low Mean Squared Error (MSE) values, spanning from 0.00966 to 0.11115, and Mean Absolute Error (MAE) values, ranging from 0.08176 to 0.11747, underscore the model’s accuracy in gross β activity prediction. The predictive capabilities of the models are robustly demonstrated, showcasing promising results. According to the results of the sensitivity analysis, the category of synoptic parameters has the most important influence on the prediction of atmospheric β radioactivity levels, namely air temperature, potential temperature and relative humidity. Regarding the residence time of the air masses, the periods spent over land or water have the most effect on gross β levels.

PREDICTION OF DRIVING TIME OF ELECTRIC SCOOTER (E-SCOOTER) DRIVERS BY MACHINE LEARNING

This study aims to estimate the driving times of drivers who prefer electric scooter vehicles. In general, e-scooters reduce the loss of time caused by traffic jams because, thanks to their smaller size and maneuverability, these vehicles provide rapid progress in urban journeys. E-scooters also offer an advantage in finding a parking space and easy parking thanks to their more compact structure. In this study, ML algorithms were used to predict the driving times of drivers who prefer e-scooter vehicles. The AB model has performed well with a low Mean Square Error (MSE) value (0.005). The Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) values are also relatively low (0.069 and 0.039, respectively), indicating that the model's predictions are close to the actual values. Also, the high R-squared-Coefficient of Determination (R2) value (0.947) suggests that this model explains the data quite well, and its predictions approach the actual values with high accuracy. On the other hand, the GB algorithm performed poorly compared to different algorithms, with its high margin of error and low accuracy rate. These results provide an advantage in time management by estimating the travel time a driver will make with the e-scooter. As a result, e-scooters offer drivers the opportunity to save time and manage their daily mobility more effectively, driving these vehicles attractive for transportation.

XGBoost machine learning assisted prediction of the mechanical and fracture properties of unvulcanized and dynamically vulcanized PP/EPDM reinforced with clay and halloysite nanoparticles

AbstractPolymer nanocomposites have found wide industrial applications, necessitating optimal mechanical and fracture properties evaluation, traditionally done through costly experimental methods. This study employs machine learning, particularly XGBoost, to predict properties like tensile and fracture properties swiftly, aiding material innovation across industries. The research investigates unvulcanized and vulcanized polypropylene (PP)/ethylene propylene diene monomer (EPDM) reinforced with clay and halloysite nanoparticles (HNT), analyzing fracture properties via essential work of fracture (EWF). Experimental design selects tests, and an XGBoost model predicts tensile strength and modulus, strain at break, EWF, and non‐EWF based on EPDM and nanoparticle percentages, composite and nanoparticle types. The model accurately predicts tensile strength and modulus but less so for strain at break, EWF, and non‐EWF. Mean Absolute Percentage Error values for training/test are 0.49/1.21, 1.05/1.55, 34.21/42.76, 3.02/14.35, and 2.89/3.78, with determination coefficients of 0.99/0.98, 0.99/0.97, 0.97/0.91, 0.97/0.79, and 0.92/0.73. Nanoparticles mainly affect outputs, with EPDM secondarily impactful, while composite and nanoparticle types exhibit similar significance. The best‐performing polymer nanocomposite is a dynamically vulcanized one containing 10 wt% EPDM and 3 wt% clay, achieving tensile strength of 25.070 MPa, tensile modulus of 261.170 MPa, EWF of 75.300 N/mm, and non‐EWF of 10.150 N/mm2.Highlights The effects of ethylene propylene diene monomer (EPDM), clay and halloysite nanoparticles on the mechanics of polypropylene‐based nanocomposites. Essential work of fracture (EWF) was used to study the fracture properties. Machine learning was employed to predict all mechanical characteristics. The vulcanization process improved all mechanical characteristics. The best compound: vulcanized one containing 10 wt% EPDM and 3 wt% clay.

Dynamic Prediction of Recidivism in Violence in Community-Based Schizophrenia Spectrum Disorder Patients: A Joint Model

To construct a model for predicting recidivism in violence in community-based schizophrenia spectrum disorder patients (SSDP) by adopting a joint modeling method. Based on the basic data on severe mental illness in Southwest China between January 2017 and June 2018, 4565 community-based SSDP with baseline violent behaviors were selected as the research subjects. We used a growth mixture model (GMM) to identify patterns of medication adherence and social functioning. We then fitted the joint model using a zero-inflated negative binomial regression model and compared it with traditional static models. Finally, we used a 10-fold training-test cross validation framework to evaluate the models' fitting and predictive performance. A total of 157 patients (3.44%) experienced recidivism in violence. Medication compliance and social functioning were fitted into four patterns. In the counting model, age, marital status, educational attainment, economic status, historical types of violence, and medication compliance patterns were predictive factors for the frequency of recidivism of violence (P<0.05). In the zero-inflated model, age, adverse drug reactions, historical types of violence, medication compliance patterns, and social functioning patterns were predictive factors for the recidivism in violence (P<0.05). For the joint model, the average value of Akaike information criterion (AIC) for the train set was 776.5±9.4, the average value of root mean squared error (RMSE) for the testing set was 0.168±0.013, and the average value of mean absolute error (MAE) for the testing set was 0.131±0.018, which were all lower than those of the traditional static models. Joint modeling is an effective statistical strategy for identifying and processing dynamic variables, exhibiting better predictive performance than that of the traditional static models. It can provide new ideas for promoting the construction of comprehensive intervention systems.

Implementing fuzzy SMRGT, ANN, and ANFIS for flow coefficient estimation in Antalya River Basin

The estimation of the flow coefficient is a vital hydrological procedure that holds considerable importance in flood prediction, water resource management, and flood mitigation. The precise estimation of the flow coefficient is imperative in mitigating flood-related damages, administering flood alert mechanisms, and regulating water discharge. It is hard to accurately determine the flow coefficient without a good understanding of the river basin’s hydrology, climate, topography, and soil characteristics. A range of methodologies have been documented in the most recent body of literaturefor flow coefficient modeling. The majority of these methods, however, depend on opaque techniques that lack generalizability. Therefore, this research employed three distinct methodologies—specifically, the Adaptive Neural Fuzzy Inference System (ANFIS), the Simple Membership Function, and the Fuzzy Rules Generation Technique (SMRGT) are all examples of fuzzy inference systems, and Artificial Neural Network (ANN), to achieve its objectives. The Aksu River Basin in Antalya, Turkey, was chosen as the study area. The models underwent multiple permutations of precipitation (P), temperature (T), relative humidity (Rh), wind speed (Ws), land use (LU), and soil properties (Sp) data that were tailored to the particular study region. The study analyzed the results using various performance metrics of the model such as mean absolute error (MAE), Nash–Sutcliffe efficiency coefficient (NSE), root mean square error (RMSE), and correlation coefficient (R2). The results indicate that the SMRGT method resulted in a remarkable degree of accuracy in forecasting the flow coefficient, as demonstrated with the minimal RMSE and MAE values and high correlation coefficient values. The study’s findings suggest that the SMRGT method was applied effectively in hydrological analysis to estimate the flow coefficient, contributing to more accurate flood prediction, water resource management, and flood mitigation strategies.

Peramalan Jumlah Kebutuhan Packaging Papper Photocopy (PPC) PT. Y Menggunakan Metode Moving Avarage dan Exponential Smoothing

In the era of globalization, industry and trade in Indonesia face fierce competition. Technological advances require quality human resources from formal and non-formal educational institutions with supporting facilities. This research focuses on solving safety stock problems using the Exponential Smoothing and Moving Average methods. Due to the current changing market demand, the results of data processing carried out using the Quality Management (QM) program for Windows show that the forecasting method with the Moving Average approach has the highest Mean Absolute Percentage Error (MAPE) value of 16.03%, while the Exponential Smoothing method has a MAPE value of 13.7%. So with this, the lowest MAPE value is in the forecasting method with the Exponential Smoothing approach of 13.7%. Keywords: Demand Forecasting, Exponential Smoothing Method, Moving Average Method, MAPE, Safety Stock

Optimizing KNN Algorithm Using Elbow Method for Predicting Voter Participation Using Fixed Voter List Data (DPT)

The purpose of this study is to produce maximum predictions for election participation rates. KNN (K-Nearest Neighbors) is one of the machine learning algorithms used to classify or regress data. The KNN algorithm works by finding the closest K training data from test data to be classified. Although the KNN (K-Nearest Neighbors) algorithm has advantages such as being easy to implement and being able to handle non-linear data, this algorithm also has several weaknesses, one of which is the determination of the value of K which is very ordinary and subjective. Therefore in this study optimization of the value of K on KNN using the Elbow method. The dataset used is the Fixed Voters List (DPT) in the 2019 General Elections in Karawang Regency. The final results of the experiments in this study, the highest achievement was obtained with a Mean Squared Error (MSE) value of 0.0018, a Root Mean Squared Error (RMSE) value of 0.0422, and a Mean Absolute Percent Error (MAPE) value of 6.36%. The highest accuracy produced in this study was 95.63% and the lowest was 93.64% with an average accuracy of 95.02%.

Green pepper fruits counting based on improved DeepSort and optimized Yolov5s.

Green pepper yield estimation is crucial for establishing harvest and storage strategies. This paper proposes an automatic counting method for green pepper fruits based on object detection and multi-object tracking algorithm. Green pepper fruits have colors similar to leaves and are often occluded by each other, posing challenges for detection. Based on the YOLOv5s, the CS_YOLOv5s model is specifically designed for green pepper fruit detection. In the CS_YOLOv5s model, a Slim-Nick combined with GSConv structure is utilized in the Neck to reduce model parameters while enhancing detection speed. Additionally, the CBAM attention mechanism is integrated into the Neck to enhance the feature perception of green peppers at various locations and enhance the feature extraction capabilities of the model. According to the test results, the CS_YOLOv5s model of mAP, Precision and Recall, and Detection time of a single image are 98.96%, 95%, 97.3%, and 6.3 ms respectively. Compared to the YOLOv5s model, the Detection time of a single image is reduced by 34.4%, while Recall and mAP values are improved. Additionally, for green pepper fruit tracking, this paper combines appearance matching algorithms and track optimization algorithms from SportsTrack to optimize the DeepSort algorithm. Considering three different scenarios of tracking, the MOTA and MOTP are stable, but the ID switch is reduced by 29.41%. Based on the CS_YOLOv5s model, the counting performance before and after DeepSort optimization is compared. For green pepper counting in videos, the optimized DeepSort algorithm achieves ACP (Average Counting Precision), MAE (Mean Absolute Error), and RMSE (Root Mean Squared Error) values of 95.33%, 3.33, and 3.74, respectively. Compared to the original algorithm, ACP increases by 7.2%, while MAE and RMSE decrease by 6.67 and 6.94, respectively. Additionally, Based on the optimized DeepSort, the fruit counting results using YOLOv5s model and CS_YOLOv5s model were compared, and the results show that using the better object detector CS_YOLOv5s has better counting accuracy and robustness.

Deep learning systems for forecasting the prices of crude oil and precious metals

Commodity markets, such as crude oil and precious metals, play a strategic role in the economic development of nations, with crude oil prices influencing geopolitical relations and the global economy. Moreover, gold and silver are argued to hedge the stock and cryptocurrency markets during market downsides. Therefore, accurate forecasting of crude oil and precious metals prices is critical. Nevertheless, due to the nonlinear nature, substantial fluctuations, and irregular cycles of crude oil and precious metals, predicting their prices is a challenging task. Our study contributes to the commodity market price forecasting literature by implementing and comparing advanced deep-learning models. We address this gap by including silver alongside gold in our analysis, offering a more comprehensive understanding of the precious metal markets. This research expands existing knowledge and provides valuable insights into predicting commodity prices. In this study, we implemented 16 deep- and machine-learning models to forecast the daily price of the West Texas Intermediate (WTI), Brent, gold, and silver markets. The employed deep-learning models are long short-term memory (LSTM), BiLSTM, gated recurrent unit (GRU), bidirectional gated recurrent units (BiGRU), T2V-BiLSTM, T2V-BiGRU, convolutional neural networks (CNN), CNN-BiLSTM, CNN-BiGRU, temporal convolutional network (TCN), TCN-BiLSTM, and TCN-BiGRU. We compared the forecasting performance of deep-learning models with the baseline random forest, LightGBM, support vector regression, and k-nearest neighborhood models using mean absolute error (MAE), mean absolute percentage error, and root mean squared error as evaluation criteria. By considering different sliding window lengths, we examine the forecasting performance of our models. Our results reveal that the TCN model outperforms the others for WTI, Brent, and silver, achieving the lowest MAE values of 1.444, 1.295, and 0.346, respectively. The BiGRU model performs best for gold, with an MAE of 15.188 using a 30-day input sequence. Furthermore, LightGBM exhibits comparable performance to TCN and is the best-performing machine-learning model overall. These findings are critical for investors, policymakers, mining companies, and governmental agencies to effectively anticipate market trends, mitigate risk, manage uncertainty, and make timely decisions and strategies regarding crude oil, gold, and silver markets.