Lower Mean Absolute Error Values Research Articles

Exploiting the synergy of SARIMA and XGBoost for spatiotemporal earthquake time series forecasting

AbstractEarthquakes are vibrations that occur on the surface of earth, generating fires, ground shaking, tsunamis, landslides and cracks. These incidents can cause severe damage and loss of life. Accurate earthquake forecasts are critical for anticipating and mitigating these hazards, which can avoid damage to buildings and infrastructure and save lives. To address the challenges given by earthquakes probabilistic nature, this paper presents a hybrid SARIMA–XGBoost approach to earthquake magnitude prediction. The suggested technique consists of a two‐step process: an exploration phase that uses exploratory data analysis, which includes descriptive statistics and data visualisation, and a prediction phase that focusses on forecasting future earthquakes. Using a large significant earthquake dataset spanning 1965–2023, the study intends to gain insights and lessons for more effective earthquake prediction methods. Further, in a comparison analysis, the results of SARIMA‐XGBoost model are compared to those of traditional ARIMA and SARIMA models. The results highlight the superior performance of the hybrid SARIMA–XGBoost model, showcasing a mean absolute error (MAE) of 0.038, a mean squared error (MSE) of 0.0040, and a root mean squared error (RMSE) of 0.068. These metrics collectively underscore the model's enhanced accuracy in forecasting earthquake magnitudes. The notably low values of MAE, MSE and RMSE indicate that our hybrid approach significantly improves prediction accuracy compared to alternative models. By integrating SARIMA's time series (TS) analysis with XGBoost's machine learning (ML) capabilities, the hybrid model reduces forecasting errors more effectively, demonstrating its clear advantage in precision.

ANN-based swarm intelligence for predicting expansive soil swell pressure and compression strength

This research suggests a robust integration of artificial neural networks (ANN) for predicting swell pressure and the unconfined compression strength of expansive soils (PsUCS-ES). Four novel ANN-based models, namely ANN-PSO (i.e., particle swarm optimization), ANN-GWO (i.e., grey wolf optimization), ANN-SMA (i.e., slime mould algorithm) alongside ANN-MPA (i.e., marine predators’ algorithm) were deployed to assess the PsUCS-ES. The models were trained using the nine most influential parameters affecting PsUCS-ES, collected from a broader range of 145 published papers. The observed results were compared with the predictions made by the ANN-based metaheuristics models. The efficacy of all these formulated models was evaluated by utilizing mean absolute error (MAE), Nash–Sutcliffe (NS) efficiency, performance index ρ, regression coefficient (R2), root mean square error (RMSE), ratio of RMSE to standard deviation of actual observations (RSR), variance account for (VAF), Willmott’s index of agreement (WI), and weighted mean absolute percentage error (WMAPE). All the developed models for Ps-ES had an R significantly > 0.8 for the overall dataset. However, ANN-MPA excelled in yielding high R values for training dataset (TrD), testing dataset (TsD), and validation dataset (VdD). This model also exhibited the lowest MAE of 5.63%, 5.68%, and 5.48% for TrD, TsD, and VdD, respectively. The results of the UCS model’s performance revealed that R exceeded 0.9 in the TrD. However, R decreased for TsD and VdD. Also, the ANN-MPA model yielded higher R values (0.89, 0.93, and 0.94) and comparatively low MAE values (5.11%, 5.67, and 3.61%) in the case of PSO, GWO, and SMA, respectively. The UCS models witnessed an overfitting problem because the aforementioned R values of the metaheuristics were 0.62, 0.56, and 0.58 (TsD), respectively. On the contrary, no significant observation was recorded in the VdD of UCS models. All the ANN-base models were also tested using the a-20 index. For all the formulated models, maximum points were recorded to lie within ± 20% error. The results of sensitivity as well as monotonicity analyses depicted trending results that corroborate the existing literature. Therefore, it can be inferred that the recently built swarm-based ANN models, particularly ANN-MPA, can solve the complexities of tuning the hyperparameters of the ANN-predicted PsUCS-ES that can be replicated in practical scenarios of geoenvironmental engineering.

Forecasting the power generation at renewable power plants in Sri Lanka using regression trees

This paper presents the application of regression trees as a versatile alternative to other machine learning and statistical modelling techniques to forecast the power generation at five renewable power plants: one large hydropower plant, two mini hydropower plants, and two wind farms in Sri Lanka. The prediction models for each power station were developed by varying the depth of the regression tree. The regression tree model with the lowest depth that forecasts the output (power) in terms of all the predictor variables was selected for each power station and the accuracy of the models was evaluated by means of the Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE), and the Coefficient of Determination (R2). According to the degree of the above performance indicators, i.e. very low values of MAE, MAPE, and RMSE supplemented by R2 of 0.95 or more, the regression tree method proved to be a convenient forecasting technique to predict the power generation at both hydro and wind power plants. Further, it could be found that a good correlation between the input and output variables paves the way for a smaller depth in the regression tree. Moreover, regression trees presented here could accurately identify the relationship between the power generated and the most influential weather factors, without being affected by potential outliers or missing values while managing collinearity too. Extension of this study would enable to generalize the prediction of renewable power generation based on the regression tree method, leading towards minimizing the use of fossil fuel.

Artificial Neural Network for Forecasting Reference Evapotranspiration in Semi-Arid Bioclimatic Regions

A correct determination of irrigation water requirements necessitates an adequate estimation of reference evapotranspiration (ETo). In this study, monthly ETo is estimated using artificial neural network (ANN) models. Eleven combinations of long-term average monthly climatic data of air temperature (min and max), wind speed (WS), relative humidity (RH), and solar radiation (SR) recorded at nine different weather stations in Tunisia are used as inputs to the ANN models to calculate ETo given by the FAO-56 PM (Penman–Monteith) equation. This research study proposes to: (i) compare the FAO-24 BC, Riou, and Turc equations with the universal PM equation for estimating ETo; (ii) compare the PM method with the ANN technique; (iii) determine the meteorological parameters with the greatest impact on ETo prediction; and (iv) determine how accurate the ANN technique is in estimating ETo using data from nearby weather stations and compare it to the PM method. Four statistical criteria were used to evaluate the model’s predictive quality: the determination coefficient (R2), the index of agreement (d), the root mean square error (RMSE), and the mean absolute error (MAE). It is quite evident that the Blaney–Criddle, Riou, and Turc equations underestimate or overestimate the ETo values when compared to the PM method. Values of ETo underestimation ranged from 1.9% to 66.1%, while values of overestimation varied from 0.9% to 25.0%. The comparisons revealed that the ANN technique could be adeptly utilized to model ETo using the available meteorological data. Generally, the ANN technique performs better on the estimates of ETo than the conventional equations studied. Among the meteorological parameters considered, maximum temperature was identified as the most significant climatic parameter in ETo modeling, reaching values of R and d of 0.936 and 0.983, respectively. The research showed that trained ANNs could be used to yield ETo estimates using new data from nearby stations not included in the training process, reaching high average values of R and d values of 0.992 and 0.997, respectively. Very low values of MAE (0.233 mm day−1) and RMSE (0.326 mm day−1) were also obtained.

Advancing heat transfer modeling through machine learning: A focus on forced convection with nanoparticles

This study underscores the critical role of heat transfer modeling in the design and optimization of heat exchangers, recognizing the challenges posed by traditional numerical methods. To address the need for more efficient alternatives, machine learning emerges as a promising solution, leveraging its capacity to unravel intricate relationships in heat transfer processes. The research focuses on forced convection heat transfer involving nanoparticles (specifically TiO2, Al2O3, and Cu) dispersed in water, employing the Lagrangian–Eulerian approach under turbulent flow conditions with a constant surface heat flux. Machine-learning techniques, namely extreme gradient boosting, support vector regression, and tree regression are employed to predict heat transfer outcomes based on collected data. The achieved results exhibit promising accuracy levels, with eXtreme gradient boosting, support vector regression, and tree regression algorithm models demonstrating impressive accuracies of 91%, 91%, and 94%, respectively. The mean absolute errors for these machine-learning models were recorded as 1.07, 7.24, and 3.74. This research serves as a demonstration of the potential of machine-learning methodologies in modeling complex physical systems, as evidenced by their efficiency in capturing the behavior of nanofluids under various conditions. The obtained accuracy and low mean absolute error values indicate the robustness of these machine-learning models in predicting heat transfer results. The implications of these findings are substantial for the field of heat exchanger design and optimization. The study suggests that machine-learning techniques offer a more efficient and cost-effective approach compared to traditional numerical simulations. By leveraging the capabilities of these models, there is potential for enhanced accuracy and efficiency in the design and optimization processes of heat exchangers, contributing to advancements in thermal engineering.

Study on the Mechanical Properties of Polyurethane-Cement Mortar Containing Nanosilica: RSM and Machine Learning Approach

Polymer-modified cement mortar has been increasingly used as a runway/road pavement repair material due to its improved bending strength, bonding strength, and wear resistance. The flexural strength of polyurethane–cement mortar (PUCM) is critical in achieving a desirable maintenance effect. This study aims to evaluate and optimize the flexural strength of PUCM involving nano silica (NS) using a central composite design/response surface methodology (CCD/RSM) to design and establish statistical models. The PU binder and NS were utilized as input parameters to evaluate the responses, such as compressive and flexural strength. Moreover, machine learning (ML) algorithms including artificial neural networks (ANN) and Gaussian regression process (GPR) were used. The PUCM mixtures were prepared by adding a PU binder at 0%, 10%, 15%, and 25% by weight of cement. At the same time, NS was incorporated into the mortar mixes at 0 to 3% (interval of 1%) by cement weight. The results showed that the simultaneous effect of PU binder at the optimal content and NS improved the performance of PUCM. Adding NS to the mortar mixture mitigated some of the strength lost due to the PU binder, which remarkably reduces the strength properties at a high content. The optimized PUCM can be obtained by partly adding 3.5% PU binder and 2.93% NS particles by the weight of cement. The performance of the machine learning algorithms was tested using performance indicators such as the determination of coefficient (R2), mean absolute error (MAE), mean-square error (MSE), and root-mean-square error (RMSE). The GPR algorithm outperformed the ANN with higher R2 and lower MAE values in the training and testing phases. The GPR can predict flexural strength with 90% accuracy, while ANN can predict it with 75% accuracy.

ARIMA approach for temperature and rainfall time series prediction in Punjab

The present study aims to explore the effectiveness of Seasonal Autoregressive Integrated Moving Average (SARIMA) models in forecasting meteorological time series data exhibiting seasonal patterns. We compared the performance of SARIMA models with different configurations and evaluate their forecasting accuracy using real-world meteorological datasetsfor three different agroclimatic zones of Punjab (sub mountainous region, central region and south west region) was analyzed to forecast mean monthly maximum air temperature, minimum air temperature and rainfall. The weather data was used from 1984-2022 for sub-mountainous zone (Ballowal Saunkhri), 1970-2022 for Central zone (Ludhiana) and 1977-2022 for south west zone (Bathinda). The results provide insights into the suitability and limitations of SARIMA models for meteorological forecasting and offer practical recommendations for practitioners and researchers in the field. The goodness of fit was tested against residuals using Ljung-Box test. The accuracy of the model was tested using Mean Absolute Error (MAE) and root square mean error (RMSE). The model achieved Mean Absolute Errors (MAE) ranging from 0.61 to 0.78 for maximum temperature, 0.74 to 0.49 for minimum temperature, and 32.12 to 45.44 for rainfall, with lower MAE values indicating higher predictive accuracy. The fitted model was able to capture dynamics of the temperature time series and produce a sensible forecast. However, the model was unable to forecast rainfall series efficiently.

Decision Tree-Based Modeling of the Aeration Effectiveness of Circular Plunging Jets.

Since soft computing has gained a lot of attention in hydrological studies, this study focuses on predicting aeration efficiency (E20) using circular plunging jets employing soft computing techniques such as reduced error pruning tree (REPTree), random forest (RF), and M5P. The study undertaken required the development and validation of models, which were achieved using 63 experimental data values with input variables, such as angle of inclination of tilt channel (α), number of plunging jets (JN), discharge of each jet (Q), hydraulic radius of each jet (HR), and Froude number (Fr. No), to evaluate the aeration efficiency (E20), which served as the output variable. To evaluate the effectiveness of the developed models, three different statistical indices were used such as the coefficient of correlation (CC), root-mean-square error (RMSE), and mean absolute error (MAE), and it was found that all of the applied techniques possessed good forecasting ability since their correlation coefficient values were greater than 0.8. Upon testing, it was discovered that the M5P model outperformed other soft computing-based models in its ability to predict E20, as demonstrated by its correlation coefficient value of 0.9564 and notably low values of MAE (0.0143) and RMSE (0.0193).

Development and evaluation of temperature-based deep learning models to estimate reference evapotranspiration

Efficient irrigation management of urban landscapes is critical in arid/semi-arid environments and depends on the reliable estimation of reference evapotranspiration (ETo). However, the available measured climatic data in urban areas are typically insufficient to use the standard Penman-Monteith for ETo estimation. Therefore, smart landscape irrigation controllers often use temperature-based ETo models for autonomous irrigation scheduling. This study focuses on developing deep learning temperature-based ETo models and comparing their performance with widely used empirical temperature-based models, including FAO Blaney & Criddle (BC), and Hargreaves & Samani (HS). We also developed a simple free and easy-to-access tool called DeepET for ETo estimation using the best-performing deep learning models developed in this study. Four artificial neural network (ANN) models were developed using raw weather data as inputs and the reconstructed signal obtained from the wavelet transform as inputs. In addition, long short-term memory (LSTM) recurrent neural network (NN) and one-dimensional convolution neural network (CNN) models were developed. A total of 101 active California Irrigation Management Information System (CIMIS) weather stations were selected for this study, with >725,000 data points expanding from 1985 to 2019. The performance of the models was evaluated against the standard CIMIS ETo. When evaluated at the independent sites, the temperature-based DL (Deep Learning) models showed 15–20% lower mean absolute error values than the calibrated HS model. No improvement in the performance of the ANN models was observed using reconstructed signals obtained from the wavelet transform. Our study suggests that DL models offer a promising alternative for more accurate estimations of ETo in urban areas using only temperature as input. The DeepET can be accessed from the Haghverdi Water Management Group website: http://www. ucrwater.com/software-and-tools.html.

Dragonfly algorithm–support vector machine approach for prediction the optical properties of blood

Knowledge of the optical properties of blood plays important role in medical diagnostics and therapeutic applications in laser medicine. In this paper, we present a very rapid and accurate artificial intelligent approach using Dragonfly Algorithm/Support Vector Machine models to estimate the optical properties of blood, specifically the absorption coefficient, and the scattering coefficient using key parameters such as wavelength (nm), hematocrit percentage (%), and saturation of oxygen (%), in building very highly accurate Dragonfly Algorithm-Support Vector Regression models (DA-SVR). 1000 training and testing sets were selected in the wavelength range of 250-1200 nm and the hematocrit of 0-100%. The performance of the proposed method is characterized by high accuracy indicated in the correlation coefficients (R) of 0.9994 and 0.9957 for absorption and scattering coefficients, respectively. In addition, the root mean squared error values (RMSE) of 0.972 and 2.9193, as well as low mean absolute error values (MAE) of 0.2173 and 0.2423, this result showed a strong match with the experimental data. The models can be used to accurately predict the absorption and scattering coefficients of blood, and provide a reliable reference for future studies on the optical properties of human blood.

Predictive Modeling of Yarn Quality at Ring Spinning Machine using Resilient Back Propagation Neural Networks

The final attenuation and twisting of fiber take place at ring spinning machine and hence its optimized performance is very crucial in terms of yarn quality. Drafting at ring spinning machine has a decisive effect on quality. There exist many influencing parameters in the spinning geometry that have to be optimized for manufacturing of quality yarn. The present research work was carried out to develop the Artificial neural networks (ANN) based prediction model for the polyester/cotton blended ring spun yarns by using these influencing parameters as inputs. ANN prediction model was developed using resilient backpropogation algorithm. Yarn quality parameters like yarn evenness, hairiness and tensile parameters were predicted. The low mean absolute error values for the yarn quality parameters proved that it is possible to predict the yarn quality on the basis of spinning geometry for cotton/polyester blended ring spun yarns using Resilient Back Propogation Neural Networks.

Machine learning approach for predicting concrete compressive, splitting tensile, and flexural strength with waste foundry sand

The scarcity of landfilling and the growing expense of disposal, recycling, and reusing industrial byproducts have become attractive alternatives to removal. There are several sorts of industrial waste and byproducts; one is waste foundry sand (WFS). Using such materials in concrete not only saves costs but also helps reduce disposal concerns, including a lower impact on natural resources, a lower CO2 impact, and enhanced mechanical and durability attributes. Concrete compressive, tensile, and flexural strengths are essential in constructing concrete structures. To identify the issue above, this research aims to develop systematic multiscale models to forecast the substantial compressive, tensile, and flexural strengths of concrete replacement with WFS for usage in the building sector with no theoretical constraints. Experimental data (137 tested) from various academic research investigations were statically examined and modeled for that objective. Artificial neural networks (ANN), gaussian processes (GP), support vector machines (SVM), and the M5P model techniques were applied for the measures. The most important aspects influencing the strength of concrete in the process model are water/cement ratio (w/c) ratio, fine aggregates, cement, coarse aggregates, WFS (%), sand, WFS (mix with others), curing days, and water. Using various modeling techniques, the compressive, tensile, and flexural strengths of concrete containing WFS can be predicted in terms of (w/c), fine aggregates, coarse aggregates, cement, sand, WFS (%), and curing time using the coefficient of determination (R2) and the root mean square error. Among the methodologies tested and based on the training data set, the models based on ANN, GP, SVM, and M5P model appear to be the most efficient. The findings of this study indicate that the ANN-based model outperforms other applied models, provided training and testing datasets are used. Based on the results of the models and statistical assessments such as coefficient of determination (R2), mean absolute error (MAE), and root means square error (RMSE), the models SVM, ANN, GP, and M5P model, respectively, predicted the compressive, tensile, and flexural strengths of concrete with WFS very well with a high value of R2 and low values of MAE and RMSE. The sensitivity investigation concludes that the curing time is the most important parameter for predicting the CS of concrete with the replacement of WFS using this data set.

Standard vs total keratometry for intraocular lens power calculation in cataract surgery combined with DMEK

To compare the prediction accuracy of standard keratometry (K) and total keratometry (TK) for intraocular lens (IOL) power calculation in eyes undergoing combined cataract surgery and Descemet membrane endothelial keratoplasty (triple DMEK). Tertiary care academic referral center. Retrospective case series. Review of 83 eyes (63 patients) that underwent triple DMEK between 2019 and 2021. Biometry measurements were obtained using a swept-source optical biometer (IOLMaster 700). 63 eyes were used for statistical analysis. Mean error, mean absolute error (MAE), SD, median absolute error, maximum absolute error, root mean squared prediction error, and the percentage of eyes within prediction errors of ±0.50 diopters (D) and ±1.00 D were calculated for 9 multivariate and third-generation formulas using K and TK values (Barrett Universal II, Yeo EVO 2.0, Cooke K6, Kane, Pearl-DGS, Haigis, Holladay 1, Hoffer Q, and SRK/T). Formulas were additionally tested by using the prediction for an IOL power 1 D below the IOL used (IOLup1D). For all formulas, MAE was lower for K than for TK by an average of 0.21 D. The lowest MAE value observed was 0.67 D for "adjusted" SRK/T using K, and the highest MAE values observed were 1.24 D and 1.24 D for nonadjusted Hoffer Q and Haigis using TK, respectively. Overall, lower MAE values were observed for multivariate formulas and SRK/T. In triple DMEK eyes, the prediction accuracy of K was higher than that of TK. The most accurate formulas were SRK/T and multivariate formulas using K with the IOLup1D adjustment.

Deep learning estimation of three-dimensional left atrial shape from two-chamber and four-chamber cardiac long axis views.

Left atrial volume is commonly estimated using the bi-plane area-length method from two-chamber (2CH) and four-chamber (4CH) long axes views. However, this can be inaccurate due to a violation of geometric assumptions. We aimed to develop a deep learning neural network to infer 3D left atrial shape, volume and surface area from 2CH and 4CH views. A 3D UNet was trained and tested using 2CH and 4CH segmentations generated from 3D coronary computed tomography angiography (CCTA) segmentations (n = 1700, with 1400/100/200 cases for training/validating/testing). An independent test dataset from another institution was also evaluated, using cardiac magnetic resonance (CMR) 2CH and 4CH segmentations as input and 3D CCTA segmentations as the ground truth (n = 20). For the 200 test cases generated from CCTA, the network achieved a mean Dice score value of 93.7%, showing excellent 3D shape reconstruction from two views compared with the 3D segmentation Dice of 97.4%. The network also showed significantly lower mean absolute error values of 3.5 mL/4.9 cm2 for LA volume/surface area respectively compared to the area-length method errors of 13.0 mL/34.1 cm2 respectively (P < 0.05 for both). For the independent CMR test set, the network achieved accurate 3D shape estimation (mean Dice score value of 87.4%), and a mean absolute error values of 6.0 mL/5.7 cm2 for left atrial volume/surface area respectively, significantly less than the area-length method errors of 14.2 mL/19.3 cm2 respectively (P < 0.05 for both). Compared to the bi-plane area-length method, the network showed higher accuracy and robustness for both volume and surface area.

Examining parking occupancy prediction models: a comparative analysis

Parking is a persistent challenge in urban areas, especially on college campuses, as limited parking spaces and increasing vehicular traffic are at the forefront of campus administrators’ concerns. Lack of parking availability leads to frustration among students, faculty, and staff and results in increased traffic congestion, environmental pollution, and reduced overall productivity. Accurate prediction of parking occupancy plays a vital role in optimizing parking resource allocation, minimizing congestion, and improving the overall campus experience. Thus, this study comprehensively analyzes parking occupancy predictions for a college campus garage, using several different models. The dataset spans from January 2022 to March 2023, providing insights into the models' performance for more than one year. The research aims to utilize random forest, decision tree, linear regression, and support vector regression (SVR) models and compare their effectiveness, using evaluation metrics such as root mean square error (RMSE), mean absolute error (MAE), and R-squared (R2). Random forest outperformed decision tree, SVR and linear regression models in terms of accuracy, as indicated by the lower MAE values. Additionally, random forest achieved a better classification performance, and a higher R2 suggests a stronger correlation between the predicted and actual occupancy values. Conversely, SVR displayed the weakest performance among the four models, with a high MAE and negative R2 value. The findings of this research will aid in improving parking management systems and contribute to the development of efficient and sustainable parking solutions on college campuses and in urban areas.

E-Learning Course Recommender System Using Collaborative Filtering Models

e-Learning is a sought-after option for learners during pandemic situations. In e-Learning platforms, there are many courses available, and the user needs to select the best option for them. Thus, recommender systems play an important role to provide better automation services to users in making course choices. It makes recommendations for users in selecting the desired option based on their preferences. This system can use machine intelligence (MI)-based techniques to carry out the recommendation mechanism. Based on the preferences and history, this system is able to know what the users like most. In this work, a recommender system is proposed using the collaborative filtering mechanism for e-Learning course recommendation. This work is focused on MI-based models such as K-nearest neighbor (KNN), Singular Value Decomposition (SVD) and neural network–based collaborative filtering (NCF) models. Here, one lakh of Coursera’s course review dataset is taken from Kaggle for analysis. The proposed work can help learners to select the e-Learning courses as per their preferences. This work is implemented using Python language. The performance of these models is evaluated using performance metrics such as hit rate (HR), average reciprocal hit ranking (ARHR) and mean absolute error (MAE). From the results, it is observed that KNN is able to perform better in terms of higher HR and ARHR and lower MAE values as compared to other models.

LiDAR modeling to determine the height of shade canopy tree in cocoa agrosystems as available habitat for wildlife

Agrosystems have different canopy strata due to shade trees that serve as available habitats for endangered species such as birds, reptiles, and mammals. LiDAR is a technology used to assess habitat quality as a support for designing conservation strategies. The objective of this research was to develop a model with data derived from LiDAR to obtain the height of the shade canopy in cocoa agrosystems, as a habitat available for wildlife species. Through the data of the height of the vegetation taken in the field and the data obtained from a LiDAR point cloud, the Canopy Height Model was generated. The data from the mapping of the canopy height model of the agrosystems taken as study sites were validated using the coefficient of determination (R2), mean absolute error (MAE), and the RMSE. The mean canopy height at the study sites was 14.63, 13.84, and 13.95 m, and the results of the validation using the model predicted canopy height shows good agreement with the actual value with an R2 of 0.86, and very low values of MAE=1.88, MSE=5.64, and RMSE=2.37, which indicates that they have an acceptable degree regarding the canopy height model between the LiDAR data and the data taken in the field. Research using LiDAR provides useful information to determine the height of the canopy, in the cocoa agrosystems up to 3 strata are found, this is due to the diversity of tree species used as shade, ranging from timber, fruit, ornamental, which are used as feeding, nesting, and resting of wildlife, in the study area populations of howler monkey species that are listed as endangered by the International Union for Conservation of Nature (IUCN), in addition to other species such as bats and birds, with the presence of these species indicate that the cocoa agrosystems, serve as a habitat for a diversity of species, which is why it is important to conserve these agrosystems in the humid tropics.

Combining autoregressive integrated moving average with Long Short-Term Memory neural network and optimisation algorithms for predicting ground water level

The groundwater resources are the essential sources for irrigation and agriculture management. Forecasting groundwater levels (GWL) for the current and future periods is an essential topic of watershed management. The prediction of GWL helps prevent overexploitation. The Auto-Regressive Integrated Moving Average model (ARIMA) is a widely known linear statistical model. One of the drawbacks of the ARIMA models is that they may not capture all existing patterns, such as non-linear parts of time series. This article introduces a new hybrid model, namely the ARIMA-Long Short-Term Memory (LSTM) neural network, to capture the linear and non-linear components of a GWL time series in the Yazd-Ardekn Plain in Iran. This study applied the ARIMA-LSTM in forecasting three-, six-, and nine-month-ahead GWL. To determine the hyperparameters of the LSTM algorithm, the Salp Swarm Algorithm (SSA), sine cosine optimisation algorithm (SCOA), particle swarm optimisation algorithm (PSOA), and genetic algorithm (GA) were coupled with the LSTM model. Two different scenarios were devised to introduce new input combinations. In the first scenario, the residual values of the ARIMA model and the lagged GWL data were inserted into hybrid and standalone LSTM models for forecasting the GWL. In the second scenario, the summation of the outputs of the ARIMA and LSTM models gave the final outputs. In terms of the content of three-month-ahead GWL predictions for the second scenario, the ARIMA-LSTM-SSA produced better results than the ARIMA-LSTM-SCOA, ARIMA-LSTM-PSOA, ARIMA-LSTM-GA, ATIMA-LSTM, LSTM, and ARIMA algorithms, which had lower mean absolute error values (MAE) of 5%, 9.4%, 15%, 38%, 42%, and 47%, respectively. However, the general results indicated that an increased forecasting horizon reduced the accuracy of the models. The new hybrid ARIMA-LSTM- SSA model was highly capable of forecasting other hydrological variables for capturing non-linear and linear elements of the time series.

Structural deformation prediction model based on extreme learning machine algorithm and particle swarm optimization

In this paper, an extreme learning machine (ELM) algorithm based on particle swarm optimization (PSO) is proposed to predict structural deformation. Taking an aqueduct located in Tiantai County, Zhejiang, China, as a case study, a series of observations of the aqueduct vertical displacements and crack openings were used to train a neural network. Then, variables representing environmental factors (air temperature), hydraulic factors (water level), and aging were selected as the influence factors input into the prediction model. Finally, the proposed PSO–ELM model was used to predict the vertical deformation and crack opening of the aqueduct, and the predicted results were compared with the monitored values using four evaluation indexes: mean absolute error (MAE), mean squared error (MSE), maximum absolute error (S), and correlation coefficient (R). The prediction results obtained using the PSO–ELM model were then compared with those obtained using the evolutionary ELM, conventional ELM, back propagation neural network, long short-term memory, and multiple linear regression models. The results indicate that the proposed PSO–ELM model has an evidently superior predictive ability, with higher values of R and lower values of MAE, MSE, and S. The proposed model can therefore be confidently used to serve as a tool similar to a “weather forecast” function to predict the vertical deformation and crack openings of an aqueduct and may be employed for other structural monitoring applications as well.

Can time-dependent double hybrid density functionals accurately predict electronic excitation energies of BODIPY compounds?

The vertical excitation energies of 12 BODIPY chromophores are benchmarked via TD-DFT, using 36 functionals covering different rungs. It was found that most TD-DFT results overestimate the excitation energies, and provide mean absolute error (MAE) values larger than 0.4 eV. The dispersion-corrected, spin-component-scaled, double-hybrid (DSD) density functionals DSD-BLYP and DSD-PBEP86 are found to have the smallest MAE values of 0.119 eV and 0.124 eV, respectively. The low MAE values of the two DSD density functionals falls in the range of errors found in the wavefunction based methods that are reported to be suitable methods for such chromophores. DSD-BLYP and DSD-PBEP86 functionals also show excellent consistency (standard deviation = 0.076 eV and 0.082 eV respectively) and good predictability (linear determination coefficient R2 = 0.933 eV and 0.917 eV respectively). MAE values of the two spin-component-scaled DH functionals of vertical fluorescence calculations are also promising (MAE around 0.12 eV).