Yield Prediction Method Research Articles

Wafer Edge Yield Prediction Using a Combined Long Short-Term Memory and Feed- Forward Neural Network Model for Semiconductor Manufacturing

In semiconductor manufacturing, maintaining a high yield and ensuring accurate yield prediction are considerably important for improving productivity, customer satisfaction, and enhancing profitability. Despite its importance and merits, achieving wafer yield prediction with high quality and accuracy is challenging. In this paper, we propose a method for wafer edge yield prediction using a combined long short-term memory (LSTM) and feed-forward neural network (FFNN) model. Unlike previous research, we focus on the edge yield because of the higher yield loss at the wafer edge. The combined LSTM-FFNN model uses a dataset divided into two types according to data characteristics. Time-series data are used in the case of LSTM, and non-time-series data are fed into the FFNN. When preparing the time-series data, comprising data related to the equipment and chambers, data of different chambers do not overlap, thereby rendering them as independent entities. The proposed model outperforms other models in terms of all evaluation metrics. The coefficient of determination of the proposed combined LSTM-FFNN model is 34.14%, which is almost 13% higher than that of the other compared models on average.

Crop yield prediction method based on an improved neural network

Crop yield prediction method based on an improved neural network

Land suitability assessments for yield prediction of cassava using geospatial fuzzy expert systems and remote sensing

Cassava has the potential to be a promising crop that can adapt to changing climatic conditions in Indonesia due to its low water requirement and drought tolerance. However, inappropriate land selection decisions limit cassava yields and increase production-related costs to farmers. As a root crop, yield prediction using vegetation indices and biophysical properties is essential to maximize the yield of cassava before harvesting. Therefore, the purpose of this research was to develop a yield prediction model based on suitable areas that assess with land suitability analysis (LSA). For LSA, the priority indicators were identified using a fuzzy expert system combined with a multicriteria decision method including ecological categories. Furthermore, the yield prediction method was developed using satellite remote sensing datasets. In this analysis, Sentinel-2 datasets were collected and analyzed in SNAP® and ArcGIS® environments. The multisource database of ecological criteria for cassava production was built using the fuzzy membership function. The results showed that 42.17% of the land area was highly suitable for cassava production. Then, in the highly suitable area, the yield prediction model was developed using the vegetation indices based on Sentinel-2 datasets with 10 m resolution for the accuracy assessment. The vegetation indices were used to predict cassava growth, biophysical condition, and phenology over the growing seasons. The NDVI, SAVI, IRECI, LAI, and fAPAR were used to develop the model for predicting cassava growth. The generated models were validated using regression analysis between observed and predicted yield. As the vegetation indices, NDVI showed higher accuracy in the yield prediction model (R2 = 0.62) compared to SAVI and IRECI. Meanwhile, LAI had a higher prediction accuracy (R2 = 0.70) than other biophysical properties, fAPAR. The combined model using NDVI, SAVI, IRECI, LAI, and fAPAR reported the highest accuracy (R2 = 0.77). The ground truth data were used for the evaluation of satellite remote sensing data in the comparison between the observed and predicted yields. This developed integrated model could be implemented for the management of land allocation and yield assessment in cassava production to ensure regional food security in Indonesia.

Jujube yield prediction method combining Landsat 8 Vegetation Index and the phenological length

It is challenging to generate a time series of vegetation indices from moderate spatial resolution Landsat Thematic Mapper images (Landsat 8) for crop yield forecasting. In addition, crop yields are correlated with phenology information, especially the fruit filling days. The objectives of this study were to identify the phenology time for making a reliable jujube yield prediction, more importantly, explore an approach that used the length of phenology growth period to improve remotely sensed estimates of inter-annual variability for yields. The best time for making jujube yield prediction was found to be during the fruit filling period, showing higher correlation coefficient (r) between vegetation indices and yields. The average NDVI for 14th and 15th half-months represented a better performance for yield prediction, with a highest r value of 0.87 for NDVI, 0.82 for SAVI, 0.73 for NDWI and 0.73 for EVI, respectively. The potential of using Landsat-NDVI for jujube yield estimation, combined with the phenological length, was preliminarily proved based on 200 observations of individual jujube orchards, showing a validated R2 of 0.85, 0.80 and 0.67, RMSE of 0.61, 0.78 and 0.85 t ha−1 for 2013, 2014 and 2016, respectively. Furthermore, the phenological adjusted model was further evaluated by inter-annual official statistic data, with R2 and RMSE values ranging from 0.38 to 0.53, and 0.31 to 0.47 t ha−1, respectively. The proposed method showed better performance between years when the fruit filling days differed greatly than the leave-one-year-out method, which was verified to well fit to jujube yield monitoring and mapping two months before harvest.

Machine Learning Regression Techniques for the Silage Maize Yield Prediction Using Time-Series Images of Landsat 8 OLI

Machine learning (ML) techniques have been utilized for the crop monitoring and yield estimation/prediction using remotely sensed data. However, these methods have been investigated less for yield prediction of some crops, such as silage maize, which can be cultivated at various times in different fields of an area. Inconsistency between fields for satellite-derived normalized difference vegetation index (NDVI) temporal profiles can lead to some difficulties in yield prediction methods using time series of remotely sensed data. Therefore, this research has investigated silage maize yield prediction based on time series of NDVI dataset derived from Landsat 8 OLI. This paper employed advanced ML techniques including boosted regression tree (BRT), random forest regression (RFR), support vector regression, and Gaussian process regression (GPR) approaches and compared their performance with some proposed conventional regression methods. For this purpose, the NDVI values of all silage maize fields were averaged and integrated to produce a two-dimensional dataset for each year. The ML techniques were employed 100 times and their evaluation metrics were used to evaluate their performances and also analyze their stability. Finally, all the results of each ML technique were averaged to produce silage maize yields. The comparisons between the results of these methods indicate that the BRT technique, with the average $R$ value higher than 0.87, outperforms other ones for all years. It was followed by RFR with almost same performance as GPR technique. This research demonstrated that some advanced ML approaches can predict the silage maize yield and they are less sensitive to inconsistency of NDVI time series. The results also showed that RFR was the most stable method to predict the maize yield in 2015, while it was trained using 2013–2014 dataset.

Daily Mapping of 30 m LAI and NDVI for Grape Yield Prediction in California Vineyards

Wine grape quality and quantity are affected by vine growing conditions during critical phenological stages. Field observations of vine growth stages are too sparse to fully capture the spatial variability of vine conditions. In addition, traditional grape yield prediction methods are time consuming and require large amount grape samples. Remote sensing data provide detailed spatial and temporal information regarding vine development that is useful for vineyard management. In this study, Landsat surface reflectance products from 2013 and 2014 were used to map satellite-based Normalized Difference Vegetation Index (NDVI) and leaf area index (LAI) over two Vitis vinifera L. cv. Pinot Noir vineyards in California, USA. The spatial correlation between grape yield maps and the interpolated daily time series (LAI and NDVI) was quantified. NDVI and LAI were found to have similar performance as a predictor of spatial yield variability, providing peak correlations of 0.8 at specific times during the growing season, and the timing of this peak correlation differed for the two years of study. In addition, correlations with maximum and seasonal-cumulative vegetation indices were also evaluated, and showed slightly lower correlations with the observed yield maps. Finally, the within-season grape yield predictability was examined using a simple strategy in which the relationship between grape yield and vegetation indices were calibrated with limited ground measurements. This strategy has a strong potential to improve the accuracy and efficiency of yield estimation in comparison with traditional approaches used in the wine grape growing industry.

A comparison of two adaptive multivariate analysis methods (PLSR and ANN) for winter wheat yield forecasting using Landsat-8 OLI images

An assumption that the non-linear method is more reasonable than the linear method when canopy reflectance is used to establish the yield prediction model was proposed and tested in this study. For this purpose, partial least squares regression (PLSR) and artificial neural networks (ANN), represented linear and non-linear analysis method, were applied and compared for wheat yield prediction. Multi-period Landsat-8 OLI images were collected at two different wheat growth stages, and a field campaign was conducted to obtain grain yields at selected sampling sites in 2014. The field data were divided into a calibration database and a testing database. Using calibration data, a cross-validation concept was introduced for the PLSR and ANN model construction to prevent over-fitting. All models were tested using the test data. The ANN yield-prediction model produced R2, RMSE and RMSE% values of 0.61, 979kgha−1, and 10.38%, respectively, in the testing phase, performing better than the PLSR yield-prediction model, which produced R2, RMSE, and RMSE% values of 0.39, 1211kgha−1, and 12.84%, respectively. Non-linear method was suggested as a better method for yield prediction.

Prediction of alkaline pulping yield: equation derivation and validation

A method is described which allows prediction of alkaline pulp yield. It is based on an equation which relates the pulp yield to the cellulose content and cellulose degree of polymerization of the alkaline pulp. From the slope and intercept of this linear equation, the lignin-free pulp yield, \({\text{Y}}_{\text{T}}^{{\prime }}\), of a pulp sample can be calculated when the lignin-free cellulose content (\({\text{G}}^{{\prime }}\)) and the cellulose degree of polymerization (DP) of the pulp fiber are determined by sugar analysis and viscosimetry, respectively. In this paper the yield prediction equation is derived and validated using model celluloses and hemicelluloses. As part of the yield prediction method, a new equation was developed which corrects the cellulose degree of polymerisation obtained from the pulp intrinsic viscosity for the hemicellulose content of the pulp. The approach was successfully validated by cooking of cotton at different temperatures (160–180 °C), alkalinities (1.30–2.60 mol/L, the equivalent to an effective alkali (as Na2O) charge of 12–24 % at a L/W ratio of 4.0 L/kg), sulfidities (15 and 30 %) and anthraquinone charges (0, 0.05 and 0.1 %). The cellulose protecting effect of AQ was confirmed by the lower number of glucose units removed by alkaline peeling compared to that removed during conventional alkaline or kraft cooking.

Comparison of different methods of grapevine yield prediction in the time window between fruitset and veraison

<p style="text-align: justify;"><strong>Aim</strong>: To compare grape yield prediction methods to determine which provide the best results in terms of earliness of prediction in the growing season, accuracy and precision.</p><p style="text-align: justify;"><strong>Methods and results</strong>: The grape yields predicted by six models – one for use at fruitset (FS), two for use at <em>veraison</em> (V1 and V2), and three for use during the lag phase (LP40, LP50 and LP60) – were compared to field-measured yields. Regressions for the yield predicted by each model were constructed. The V1 and V2 models had the highest R<sup>2</sup> (0.75) and efficiency index (EF; 0.67-0.71) and the lowest RMSE values (±16-17%, or <0.5 kg per m of row). The FS model had the same or similar R<sup>2</sup> (0.58), EF (0.06) and RMSE (±30%, or <0.83 kg per m of row) values as the LP models, but allowed yield predictions to be made one month earlier.</p><p style="text-align: justify;"><strong>Conclusion</strong>: The validated FS, V1 and V2 models are all useful in predicting grape yields and could be used to accurately forecast (with different errors) grape yields at either early or later time points according to winery needs. These models could be improved as further data become available in following seasons.</p><p style="text-align: justify;"><strong>Significance and impact of the study</strong>: Few validated models are available for predicting grapevine yields at fruitset and <em>veraison</em>. This study provides predictive models that can be used at these different times of the growth cycle.<strong></strong></p>

Prediction of Standard Lactation Curves for Primiparous Holstein Cows by Using Corrected Regression Models

Prediction of the expected milk yield is important for the management of the primiparous cows (PPC) with a few or no data on their own milk productivity. We developed a system of regression equations for predicting milk yields in standard lactation. The models include the systematic effects of the calving season, the five-year rolling herd average of milk yield of PPC, the breeding values of the parents for milk production, and daily milk recordings. A total of 21,901 lactations of Holstein PPC were collected during the regular monthly milk recordings of cows in the Republic of Slovenia. By including daily milk recordings in the model, the coefficients of determination of regression models for the prediction of milk yield increase: without known recordings (M0) R2=0.80; with one recording (M1) R2=0.82; with two first consecutive recordings (M2) R2=0.86; and with three recordings (M3) R2=0.89. Deviations of milk yield up to 500 kg in a standard lactation (<1.6 kg/day) were as follows: with the model M0, they occurred in 53.4% of PPC; with M1, they occurred in 56.3% of PPC; with M2, they occurred in 64.5% of PPC; and with M3, they occurred in 70.9% of PPC. We concluded that the developed system of regression models is an appropriate method for milk yield prediction of PPC.

Combined Spectral and Spatial Modeling of Corn Yield Based on Aerial Images and Crop Surface Models Acquired with an Unmanned Aircraft System

Precision Farming (PF) management strategies are commonly based on estimations of within-field yield potential, often derived from remotely-sensed products, e.g., Vegetation Index (VI) maps. These well-established means, however, lack important information, like crop height. Combinations of VI-maps and detailed 3D Crop Surface Models (CSMs) enable advanced methods for crop yield prediction. This work utilizes an Unmanned Aircraft System (UAS) to capture standard RGB imagery datasets for corn grain yield prediction at three early- to mid-season growth stages. The imagery is processed into simple VI-orthoimages for crop/non-crop classification and 3D CSMs for crop height determination at different spatial resolutions. Three linear regression models are tested on their prediction ability using site-specific (i) unclassified mean heights, (ii) crop-classified mean heights and (iii) a combination of crop-classified mean heights with according crop coverages. The models show determination coefficients \({R}^{2}\) of up to 0.74, whereas model (iii) performs best with imagery captured at the end of stem elongation and intermediate spatial resolution (0.04m\(\cdot\)px\(^{-1}\)).Following these results, combined spectral and spatial modeling, based on aerial images and CSMs, proves to be a suitable method for mid-season corn yield prediction.

Research on Grain Yield Prediction Method Based on Improved PSO-BP

Aimed at the highly nonlinear and uncertainty of grain yield changes, a new method for grain yield prediction based on improved PSO-BP is proposed. By introducing mutation operation and adaptive adjust of inertia weight, the problem of easy to fall into local optimum, premature, low precision and low later iteration efficiency of PSO are solved. By using the improved PSO to optimize BP neural network’s parameters, the learning rate and optimization capability of conventional BP are effectively improved. The simulation results of grain production prediction show that the predict accuracy of the new method is significantly higher than that of conventional BP neural network method, and the method is effective and feasible.

Research on Grain Yield Prediction Method Based on Improved PSO-BP

Research on Grain Yield Prediction Method Based on Improved PSO-BP

Research on Grain Yield Prediction Method Based on Improved PSO-BP

Aimed at the highly nonlinear and uncertainty of grain yield changes, a new method for grain yield prediction based on improved PSO-BP is proposed. By introducing mutation operation and adaptive adjust of inertia weight, the problem of easy to fall into local optimum, premature, low precision and low later iteration efficiency of PSO are solved. By using the improved PSO to optimize BP neural network’s parameters, the learning rate and optimization capability of conventional BP are effectively improved. The simulation results of grain production prediction show that the predict accuracy of the new method is significantly higher than that of conventional BP neural network method, and the method is effective and feasible.

Assessment of cluster yield components by image analysis.

Berry weight, berry number and cluster weight are key parameters for yield estimation for wine and tablegrape industry. Current yield prediction methods are destructive, labour-demanding and time-consuming. In this work, a new methodology, based on image analysis was developed to determine cluster yield components in a fast and inexpensive way. Clusters of seven different red varieties of grapevine (Vitis vinifera L.) were photographed under laboratory conditions and their cluster yield components manually determined after image acquisition. Two algorithms based on the Canny and the logarithmic image processing approaches were tested to find the contours of the berries in the images prior to berry detection performed by means of the Hough Transform. Results were obtained in two ways: by analysing either a single image of the cluster or using four images per cluster from different orientations. The best results (R(2) between 69% and 95% in berry detection and between 65% and 97% in cluster weight estimation) were achieved using four images and the Canny algorithm. The model's capability based on image analysis to predict berry weight was 84%. The new and low-cost methodology presented here enabled the assessment of cluster yield components, saving time and providing inexpensive information in comparison with current manual methods.

Near real-time prediction of U.S. corn yields based on time-series MODIS data

Annual variation of U.S. corn production is an important matter of world concern. To assure immediate response to large-scale harvest failure in crop exporting regions, as was the case during the severe U.S. drought in 2012, and to enhance global food security, a practical crop growth monitoring system based on satellite data is required. This study developed a practical method for near real-time prediction of U.S. corn yields using Moderate Resolution Imaging Spectroradiometer (MODIS)-derived Wide Dynamic Range Vegetation Index (WDRVI) taken 7days before the corn silking stage. We incorporated two algorithms into the MODIS-based corn yield prediction method; namely, (1) a MODIS-based crop classification algorithm in consideration of differences in emergence dates between corn and soybean, and (2) a simple bias correction algorithm for correcting region-dependent yield prediction errors. The method is able to predict the annual variation of national and state level corn grain yields with high accuracy in early August and detect corn yield reductions and poor-harvest regions due to drought damage, as in 2002 and 2012, on a near real-time basis in advance of those provided by the U.S. Department of Agriculture's National Agricultural Statistics Service. The method also predicted a national-level U.S. corn grain yield for 2013, which was 3.8% lower than the NASS-statistical data.

Using Low Resolution Satellite Imagery for Yield Prediction and Yield Anomaly Detection

Low resolution satellite imagery has been extensively used for crop monitoring and yield forecasting for over 30 years and plays an important role in a growing number of operational systems. The combination of their high temporal frequency with their extended geographical coverage generally associated with low costs per area unit makes these images a convenient choice at both national and regional scales. Several qualitative and quantitative approaches can be clearly distinguished, going from the use of low resolution satellite imagery as the main predictor of final crop yield to complex crop growth models where remote sensing-derived indicators play different roles, depending on the nature of the model and on the availability of data measured on the ground. Vegetation performance anomaly detection with low resolution images continues to be a fundamental component of early warning and drought monitoring systems at the regional scale. For applications at more detailed scales, the limitations created by the mixed nature of low resolution pixels are being progressively reduced by the higher resolution offered by new sensors, while the continuity of existing systems remains crucial for ensuring the availability of long time series as needed by the majority of the yield prediction methods used today.

Yield Prediction Method of a Reaction in a Micro-Fluid Device and Experimental Proof of Increased Production Using Parallel-Connected Devices

The yields of consecutive reactions in a micro-fluid device were predicted by using Monte Carlo simulation. It was found that the yield of main product was improved by using a micro-fluid device when the ratio of the reaction rate constants is k1/k2 >1. The yield prediction map was also made by using non-dimensional Damkohler Number Da and the ratio of the reaction-rate constants k1/k2. The validation experiments of four kinds of consecutive reactions were conducted. Next, the micro-fluid system of 20 parallel-connected micro-fluid devices was developed. The maximum flow rate of the micro-fluid system was 10 mm3/s, which corresponds to 315 t/year. Evaluation of the chemical performance of the micro-fluid system was conducted using a nitration reaction. It was found that the micro-fluid system increased the production scale without decreasing the yield of the products.

A Proposed Method for Stress Analysis and Yield Prediction in Soybean Using Light Interception and Developmental Timing

Analyses of stress problems affecting soybean production would be improved if simple, rapid, easy, and cheap methods were developed for monitoring seasonal crop growth patterns. Although previous studies have established total dry matter at R1 [TDM(R1)] and TDM(R5) criteria for optimal yield, determination of these parameters in commercial soybean fields by conventional destructive methods is not practical. The objectives of the current study were to determine and validate simple regression models for estimation of TDM(R1) and TDM(R5); and to validate a simple regression model (determined from a previous study) for prediction of yield from TDM(R5). Data for development and validation of regression models were taken from 11 studies conducted at Baton Rouge, LA (1988-1998). Regression models were developed for prediction of TDM(R1) and TDM(R5) based on light interception (LI) and days to R1 and R5. The R2 values between predicted and observed TDM(R1) and predicted and observed TDM(R5) were 0.84 (P < 0.0001) and 0.94 (P < 0.0001), respectively. Predicted and observed yields were highly correlated, although R2 values were lower when several cultivars (R2 = 0.63, P < 0.0001) vs. a single cultivar (R2 = 0.84, P < 0.0001) were studied. In conclusion, it was clearly demonstrated that simple regression models based on LI and developmental timing can be developed to monitor seasonal crop growth patterns and predict yield.

Computer vision system for wild blueberry fruit yield mapping

Wild blueberry (Vaccinium angustifolium Ait.) is a unique crop native to Northeastern North America with annual production of 82,000 t from an area of 79,000 ha. Yield maps along with topography and soil nutrient maps could be used to develop precise site-specific nutrition programs for wild blueberry production. Here, the development of an automated yield monitoring and mapping system for real-time fruit yield estimation for wild blueberry is presented. The hardware components of the system consist of a digital colour camera, a positioning system receiver and a computer mounted on a specially designed farm motorised vehicle. The yield prediction method was based on the estimation of the blue pixels representing ripe fruit in the field of view of each image and their expression as a percentage of total image pixels, using it as the correlation parameter with the corresponding yield. The image processing software was development in “Delphi” and “C” programming languages. Field experiments were carried out in two commercial wild blueberry fields. Concerning the calibration of the system, the experimental results show a significant correlation between percentage blue pixels and actual fruit yield (R2 = 0.90; P