Advanced Agricultural Decision System Using Recurrent Polynomial Network for Multi-Crop Recommendation Tasks ↔ Advanced Agricultural Decision System using Recurrent Polynomial Network for Multi-Crop Recommendation Tasks

Advanced Agricultural Decision System Using Recurrent Polynomial Network for Multi-Crop Recommendation Tasks P. Prasanna1, Oggu Raju2, Gai Anjali2, Sandeboina Vineesha2, Alakuntla Venkat2 1Assistant Professor, 2UG Student, 1,2Department B. Ramesh1*, Gurrala Gangadari1, Amboju Sai Kumar1, Musham Sai Varun1, S. Praneeth Sagar1 1Department of Computer Science and Engineering (Data science)

1,2Vaagdevi Engineering College, Bollikunta, Warangal, 506005, (DS), Kommuri Pratap Reddy Institute of Technology, Ghanpur, Ghatkesar, 501301, Telangana, India India. *Correspondence: B. Ramesh Abstract Agriculture has significantly transitioned from traditional, experience-based traditional manual practices to modern datadriven data-driven approaches with the integration advancement of digital technologies and Machine Learning (ML). Earlier, crop monitoring and decision-making farmers relied heavily on manual observation, farmer expertise, experience and basic statistical techniques. However, methods, which are insufficient for handling the rapid growth increasing volume of agricultural data from sensors, weather records, and satellite imagery has created a need for intelligent systems. Traditional systems capable of handling large and complex datasets efficiently. Traditional methods often fail to process large datasets and capture intricate complex relationships among variables such as soil characteristics, properties, climate conditions, and crop health, leading to lower resulting in low prediction accuracy, delayed decisions, accuracy and inefficient resource usage. decision-making. This study addresses the limitation lack of existing systems that perform classification and regression an integrated framework capable of handling multiple agricultural prediction tasks separately, resulting in fragmented and inconsistent outputs. To overcome this, simultaneously. a unified multi-task agricultural analysis framework based on the 1CA2RT (One One Classification and Two Regression Tasks) Tasks (1CA2RT) approach is proposed. The system performs It includes crop disease classification along with NDVI and two regression tasks: Normalized Difference Vegetation Index (NDVI) estimation and harvest time prediction simultaneously. prediction. The framework employs ML models including such as Support Vector Machine (SVM)-1CA2RT, AdaBoost (AB)- 1CA2RT, and Ridge (R)-1CA2RT, along with a Deep Learning-based hybrid model Hybrid Recurrent Polynomial Ensemble (HRPE-1CA2RT) model. This (HRPE)-1CA2RT. The hybrid model integrates a Recurrent Polynomial Network (RPN) using Bidirectional Long Short-Term Memory (Bidirectional LSTM)-based Recurrent Polynomial Network (BiLSTM) with ensemble Tao Tree methods to capture complex nonlinear patterns. The system is implemented using Flask Classifier (ETTC) and SQLite for secure, real-time interaction. Ensemble Tao Tree Regressor (ETTR). Experimental results demonstrate outstanding show that HRPE-1CA2RT achieves superior performance, achieving perfect classification with 100% accuracy, precision, recall, and regression accuracy.

Key words: F1-score for classification, and an R2 score of 1.0000 for NDVI and harvest prediction. This unified framework improves prediction accuracy, consistency, and efficiency, supporting reliable decision-making in modern agriculture. Keywords: Agriculture, Data-Driven Agriculture, Machine Learning (ML), farming, Multi-Task Learning, framework, Crop Disease Classification, NDVI vegetation index Estimation, Harvest Time Prediction, 1CA2RT Framework, Hybrid Recurrent Polynomial Ensemble (HRPE), Bidirectional LSTM, Ensemble Learning. integrated prediction system, agricultural data analysis, decision support.

1. Introduction Vegetation indices have become an essential component a crucial role in precision agriculture, offering agriculture and crop monitoring by providing a straightforward and reliable assessment of the condition and quantitative means to evaluate crop health of crops [1]. Depending on the vegetation index, information on various aspects of plant growth and field variability [1]. These indices are derived from spectral information captured through remote sensing technologies and provide insights into key plant attributes development can be monitored, such as chlorophyll concentration, content, leaf area index, area, canopy structure, and water content. status, as shown in fig 1. This information can then be used to optimize prescription rates in precision agriculture, Such parameters are critical for understanding crop growth stages and detecting early signs of stress, as illustrated in Figure 1. By leveraging this information, farmers can make informed decisions regarding input management, including optimized variable fertilizer application, efficient irrigation scheduling, irrigation, and controlled pesticide usage application [2]. This targeted approach is typically implemented generally performed by identifying intrafield variability, where specific intra-field zones exhibiting stress that are underperforming or reduced performance are treated selectively experiencing stress, and target inputs to enhance overall productivity.

One of the major advantages of those areas to improve crop productivity and yield. vegetation indices is their also provide a cost-effective and non-destructive nature way of crop monitoring, ensuring a widely available and cost-effectiveness, enabling continuous monitoring without physically disturbing crops [3]. This makes them highly suitable for large-scale agricultural systems while also promoting environmentally sustainable practices. approach for assessing crop health [3]. The rapid advancement development of remote sensing technologies, including remote-sensing sensors for crop monitoring in both broadband and narrowband sensors, has significantly expanded the scope of vegetation analysis [4]. As a result, many bands open immense possibilities for their combination into novel vegetation indices have been developed, exceeding 500, each designed to capture specific crop or environmental characteristics. These indices vary in their sensitivity to different factors, such as soil background, atmospheric conditions, and crop type.

Figure. [4]. Fig. 1: Normalized difference Vegetation Index. However, To date, this has led to the abundance development of 519 total vegetation indices, per Index DataBase. While most of these indices also introduces challenges in selecting the most appropriate one for accurate crop health assessment. The performance of each index depends on multiple factors, including serve a different purpose and have unique advantages and limitations according to sensor type, type and field conditions, and crop characteristics, making the difficulty of objective evaluation difficult. assessment of their performance In agricultural production, crophealth monitoring arose. achieving maximum crop yield with minimal input at minimum cost is a primary goal. Therefore, one of the goals of agricultural production. early detection and management of problems associated with crop stress yield indicators can help increase yield and yield-related issues is crucial for timely intervention subsequent profit [5]. By enabling precise monitoring and data-driven decision-making, vegetation indices play a key role in improving crop productivity, resource efficiency, and overall farm profitability.

2. Literature Survey Nițu.A.et al. [6] investigated their distinctiveness and discriminative power in the context of applications for agriculture based on hyperspectral data. More precisely, this paper merges two complementary perspectives: an unsupervised analysis with PRISMA satellite imagery to explore whether these indices are truly distinct in practice and a supervised classification over UAV hyperspectral data. They assess their discriminative power, statistical correlations, and perceptual similarities.

Tang, H.et al. [7] employed the latest Global Inventory Modelling and Mapping Studies (GIMMS) Normalized Difference Vegetation Index (NDVI4g), an updated version succeeding GIMMS NDVI3g spanning from 1982 to 2022.They integrated this dataset with the multiple scale Standardized Precipitation Evapotranspiration Index (SPEI 1 to 24) to investigate the spatial–temporal variability of sensitivity and lag in vegetation growth in response to water variability across China.Their findings indicate that over 83% of China’s vegetation demonstrates positive sensitivity to water availability, with approximately 66% exhibiting a shorter response lag (lag < 1 month). This relationship varies across aridity gradients and diverges among plant functional types. Over 66% of China’s vegetation displays increased sensitivity to water variability and 63% manifests a short response lag to water changes over the past 41 years. These outcomes significantly contribute to understanding vegetation dynamics in response to changing water conditions, implying a heightened susceptibility of vegetation to drought in a future warming world.

Kaya, F et al. [8] investigated the impact of spatial resolution on classifying three-year, multi-temporal vegetation indices derived from satellites with coarse (30 m, Landsat 8), medium (10 m, Sentinel-2), and fine spatial resolutions (3.7 m, PlanetScope). The classification was performed using the fuzzy cmeans algorithm, with the fuzziness performance index (FPI) and normalized classification entropy (NCE), which were used to determine the optimal number of management zones (MZs). The Their results revealed that the Landsat 8-based NDVI images produced the highest number of clusters (nine for annual cropland and six for orchards), while the finer resolutions from PlanetScope reduced this to three clusters for both cultivation types, more accurately capturing the intra-parcel variability. Except for Landsat 8, the NDVI means of MZs generated based on Sentinel-2 and PlanetScope using the fuzzy cmeans algorithm showed statistically significant differences from each other, as determined by a oneway and Welch’s ANOVA (p < 0.05). Aslan et al. [9] emphasized the need for comprehensive comparisons and more consistent methodologies in future research. Their work underscores the significant role of Sentinel-2 and AI in advancing precision agriculture, offering valuable insights for future studies that aim to enhance sustainability and efficiency in crop management through advanced predictive models.

He, Q. et al. [10] highlighted a considerable and widespread greening on the LP from 1982 to 2022, evidenced by a kNDVI slope of 0.0020 yr−1 (p < 0.001) and a 90.9% significantly increased greened area. The GTGP expedited this greening process, with the kNDVI slope increasing from 0.0009 yr−1 to 0.0036 yr−1 and the significantly greened area expanding from 39.1% to 84.0%. Over the past 40 years, the LP experienced significant warming (p < 0.001), slight humidification, and a marginal decrease in SR. Post-GTGP PostGTGP implementation, the warming rate decelerated, while PRE and SR growth rates slightly accelerated. Vidican, R. et al. [11] showed that VIs appears appear to be suitable for mapping and monitoring agricultural crops, forage crops, meadows, meadows and pastures. Sentinel-1 and Sentinel-2 data were the most utilized sources, while some of the frequently used VIs were EVI, LAI, NDVI, GNDVI, PSRI, and SAVI. In most of the studies, an array of VIs needed to be employed to achieve a good discrimination of crops or prediction of yields. The main challenges in using VIs are related to the variation of the spectral characteristics during the vegetation period and to the similarities of the spectral signatures of various crops and semi-natural meadows. Thus, further studies are needed to establish appropriate models for the use of satellite data that would prove to have greater accuracy and provide more relevant information for the efficient monitoring of agricultural crops.

Robinson, N.P. et al. [12] addressed this deficiency by producing a Landsat derived, high resolution (30 m), long-term (30+ years) NDVI dataset for the conterminous United States. They used Google Earth Engine, a planetary-scale cloud-based geospatial analysis platform, for processing the Landsat data and distributing the final dataset. They used a climatology driven approach to fill missing data and validate the dataset with established remote sensing products at multiple scales. They provided access to the composites through a simple web application, allowing users to customize key parameters appropriate for their application, question, and region of interest. Krakauer, N.Y. et al. [13] analysed analyzed the NDVI normalized difference vegetation index (NDVI) from 1981 to 2015 semi-monthly, semimonthly, at an 8 km spatial resolution. They used a random forest (RF) of regression trees to generate a statistical model of the NDVI as a function of elevation, land use, CO2 level, temperature, and precipitation. They found that the NDVI increased over the studied period, particularly at low and middle elevations and during the fall (post-monsoon). They inferred from the fitted RF model that the NDVI linear trend is primarily due to CO2 level (or another environmental parameter that is changing quasi-linearly), and not primarily due to temperature or precipitation trends. On the other hand, interannual fluctuation in the NDVI is more correlated with temperature and precipitation. The RF accurately fits the available data and shows promise for estimating trends and testing hypotheses about their causes.

Zhao, Q. et al. [14] addressed the limitations and meet the needs of vegetation monitoring research and remote-sensing NDVI validation, his study implemented a novel NDVI camera. The proposed camera incorporates narrowband dual-pass filters designed to precisely separate red and near-infrared (NIR) spectral bands, which are aligned with the configuration of sensors onboard satellites. Through software-controlled imaging parameters, the camera captures the real radiance of vegetation reflection, ensuring the acquisition of accurate NDVI values while preserving the evolving trends of the vegetation status. The performance of this NDVI camera was evaluated using a hyperspectral spectrometer in the Hulunbuir Grassland over a period of 93 days. Their results demonstrated distinct seasonal characteristics in the camera-derived NDVI time series using the Green Chromatic Coordinate (GCC) index. Moreover, in comparison to the GCC index, the camera’s NDVI values exhibit greater consistency with those obtained from the hyperspectral spectrometer, with a mean deviation of 0.04, and a relative root mean square error of 9.68%. Eastman, J.R. et al. [15] used the Seasonal Trend Analysis (STA) procedure, over half (56.30%) of land surfaces were found to exhibit significant trends. Almost half (46.10%) of the significant trends belonged to three classes of seasonal trends (or changes). Class 1 consisted of areas that experienced a uniform increase in NDVI throughout the year, and was primarily associated with forested areas, particularly broadleaf forests. Class 2 consisted of areas experiencing an increase in the amplitude of the annual seasonal signal whereby increases in NDVI in the green season were balanced by decreases in the brown season. These areas were found primarily in grassland and shrubland regions. Class 3 was found primarily in the Taiga and Tundra biomes and exhibited increases in the annual summer peak in NDVI. While no single attribution of cause could be determined for each of these classes, it was evident that they are primarily found in natural areas (as opposed to anthropogenic land cover conversions) and that they are consistent with climate-related ameliorations of growing conditions during the study period.

3. Proposed System The methodology establishes a structured and data-driven approach for intelligent agricultural analysis by integrating multiple prediction tasks within a unified framework. It follows a systematic pipeline that begins with data acquisition from agricultural datasets, followed by preprocessing and feature extraction to prepare the data for analytical modelling. The framework supports simultaneous execution of one classification task and two regression tasks, enabling comprehensive analysis of crop conditions, vegetation health, and harvest timelines. Multiple ML models are employed to learn patterns from the data, while a hybrid learning mechanism enhances prediction performance and adaptability, as presented in fig. 2. A lightweight database system ensures efficient storage and retrieval of user and prediction data, and a web-based interface enables real-time interaction, visualization, and monitoring. Continuous model evaluation and retraining further improve system performance, allowing it to adapt to changing agricultural conditions and maintain long-term accuracy. Fig. 2: proposed system architecture presents a unified pipeline for multi-task agricultural prediction by integrating data processing, hybrid model training, and deployment. Initially, raw agricultural data is collected and pre-processed to extract key features related to soil, weather, and crop conditions. Architecture. User Interface (Web Browser) ● The user interacts with The system employs multiple models including SVM-1CA2RT, AB-1CA2RT, R-1CA2RT, through a browser-based graphical interface. ● Users can perform operations such as registration, login, data input, single prediction, batch prediction, and viewing analytical results. ● The interface allows visualization of EDA plots, model performance metrics, and prediction outputs. ● All interactions are converted into HTTP requests and sent to the Flask server. Flask Web Server (app.py) ● The Flask backend receives user requests and routes them to appropriate modules. ● It manages authentication, session handling, prediction workflows, and data visualization. ● The server coordinates communication between the user interface, ML models, and the proposed HRPE-1CA2RT database. ● It also handles model loading, saving, and triggering retraining when required. SQLite Database (agriculture.db) ● The database stores persistent information such as user credentials and registration details. ● It ensures secure storage of user data using encrypted passwords. ● The Flask server interacts with the database for performing classification authentication and regression tasks. The HRPE1CA2RT combines RPN (Bidirectional LSTM with attention) data retrieval. ● It provides efficient and ETT models (ETTC and ETTR) to capture both nonlinear patterns and interpretable rules. A competition-based selection mechanism chooses the best-performing model lightweight data management for each task. The system predicts crop disease status, NDVI values, and harvest time, and stores trained models using Joblib. The architecture is deployed using Flask, while SQLite is used for managing user and engineer data, enabling secure access and real-time interaction, as illustrated in Figure 2.

Data Collection and Preprocessing: The system begins by collecting operations. raw data from CSV files containing (CSV Input) ● The agricultural attributes dataset serves as the primary input source for analysis. ● It contains features such as soil moisture, soil pH, temperature, rainfall, humidity, sunlight hours, pesticide usage, yield, latitude, and pesticide usage. longitude. ● This Data cleaning is performed fed into the preprocessing pipeline for further analysis. Data Preprocessing & Feature Extraction (ml_models.py) ● The dataset undergoes cleaning, normalization, and transformation to handle missing values and inconsistencies using Pandas. Relevant ensure high-quality inputs. ● Numeric features are selected standardized using scaling techniques to ensure that only important parameters maintain feature parity. ● Irrelevant or inconsistent data is handled to prevent noise from affecting model performance. ● Feature vectors are used generated and passed to the various ML models. Existing Baseline Models (SVM-1CA2RT, AB-1CA2RT, R-1CA2RT) The processed feature set is input to baseline models for model training.

Data Splitting and Normalization: The dataset is divided into training and testing sets using an 80:20 ratio to evaluate model performance effectively. Standard Scaler normalization is applied to standardize feature values and improve model convergence. This step ensures that all models receive uniformly scaled input data.

Figure. 2: Proposed system architecture. Model Training (ML and Hybrid Models): The system trains multiple models including SVM, AdaBoost, and Ridge for benchmarking: ● SVM-1CA2RT: Handles both classification and regression tasks. In addition, ● AB-1CA2RT: Utilizes ensemble-based learning for improved prediction stability. ● R-1CA2RT: A linear model with regularization for enhanced stability against multicollinearity. ● These models independently generate predictions for Crop Disease Status, NDVI Index, and Days to Harvest. Proposed Hybrid HRPE-1CA2RT Model This is the proposed HRPE core intelligent model is trained, which integrates of the system, combining deep learning and rule-based approaches: 1. RPN: Processes feature vectors using Bidirectional LSTM layers to capture complex temporal and nonlinear relationships. It applies polynomial attention mechanisms to enhance the importance of specific agricultural features. 2. ETT: Uses decision tree-based learning for classification and regression, providing interpretable rule-based predictions that enhance model transparency. Selection Logic (Best Model Selection) ● The system performs a real-time comparison between the outputs of the RPN (Bidirectional LSTM with polynomial attention) and ETT models (ETTC and ETTR). This combination allows models. ● the system to learn complex patterns while maintaining interpretability.

Competition-Based Model Selection: All trained models are evaluated using with the superior performance metrics such as accuracy, precision, recall, F1-score, and (measured via accuracy for classification or R2 score. A competition-based mechanism selects the best-performing model score for each task. regression) is dynamically selected. ● This ensures that the most accurate final output benefits from optimal prediction accuracy and efficient model is used decision reliability. Prediction Results & Target Output ● The system generates precise predictions for prediction.

Multi-Task Prediction: The selected models are used to perform three tasks simultaneously: essential agricultural targets: o Target 1: crop disease classification (N, L, M, H), Status o Target 2: NDVI regression, and Index o Target 3: days to harvest prediction. This unified approach avoids ● Results and model-wise comparisons are displayed in an intuitive format on the need for separate systems user interface. Model Retraining Mechanism ● The system supports an adaptive learning cycle using updated agricultural datasets. ● New data is processed through the standard pipeline, and ensures consistent outputs across tasks.

Model Storage models are retrained to improve performance and Deployment: The trained adaptability. ● Updated models and scalers are saved using joblib for reuse without retraining. securely, ensuring The system is deployed using Flask, providing interfaces for agricultural engineers includes EDA evolves alongside changing crop cycles and model weather patterns. 4. RESULTS analysis and users can perform single and batch prediction). SQLite is used as the backend database to securely store user credentials and manage authentication.

4. Result Description This section presents the experimental results and interface outputs of the proposed hybrid predictive system. The figures illustrate the implementation workflow, system interfaces, model performance, and prediction outputs. The results demonstrate the effectiveness of the hybrid architecture combining DL deep learning and ensemble models. A comparative evaluation with traditional models highlights the superiority of the proposed RPN-ETT 1CA2RT CART framework.

Figure. 3: Presents performance metrics of proposed HRPE 1CA2RT model Figure Fig. 3 presents the performance results of the proposed hybrid HRPE 1CA2RT model. The figure highlights superior accuracy, reduced error rates, and improved robustness compared to all baseline models. The integration of Recurrent Polynomial Networks with Bidirectional LSTM enhances deep feature extraction from NDVI and temporal data. The Ensemble Tao Tree framework strengthens classification and regression stability. The visualization confirms the effectiveness of the hybrid architecture in precision agriculture prediction.

Figure. Fig. 3: Presents performance metrics of proposed HRPE 1CA2RT model Fig. 4: Presents model performance comparison Fig. 4 provides a comparative analysis of all evaluated models including SVM-1CA2RT, AB-1CA2RT, R-1CA2RT, and the proposed HRPE-1CA2RT model. The comparison is displayed through bar charts or performance tables. The proposed model achieves the highest accuracy and lowest error metrics across all evaluation parameters. The visual comparison clearly demonstrates the advantages of hybrid deep learning and ensemble integration. This figure validates the research objective of improving prediction reliability. Fig. 5: Presents the predictions screen Figure 4 Fig. 5 displays the final prediction interface of the system. This screen allows users to input environmental parameters such as soil conditions, climate values, and NDVI indicators. The hybrid RPN-ETT model processes the input and generates outputs including crop disease classification, NDVI estimation, and harvest timeline prediction. The results are displayed in a clear and interpretable format. This figure demonstrates the practical usability of the deployed intelligent agriculture system. 4.1 Comparative Analysis Table 1 presents the classification performance of different models for predicting crop disease status using agricultural and NDVI features. The Support Vector model achieves strong performance with high accuracy and ROC AUC values. AB and Ridge models show reduced classification effectiveness due to weaker decision boundaries and limited generalization. The Proposed HRPE Classifier achieves perfect classification across all metrics, demonstrating superior learning capability and optimized ensemble decision-making. The results confirm that the hybrid DL deep learning and ensemble architecture effectively captures nonlinear agricultural patterns and produces highly accurate disease predictions.

Table. Table 1: Classification Results – Crop Disease Status Model Name Accuracy Precision Recall F1-Score ROC AUC SVM 0.9100 0.9106 0.9100 0.9101 0.9878 AB 0.4650 0.4802 0.4650 0.4618 0.7096 Ridge 0.3267 0.3278 0.3267 0.3269 N/A HRPE 1.0000 1.0000 1.0000 1.0000 N/A Table 2 evaluates regression models used to predict NDVI index values, which represent crop vegetation health. The Support Vector regressor produces stable performance with lower error values and higher R2 score compared to other traditional methods. AB and Ridge regressors produce higher prediction errors and reduced variance explanation. The Proposed HRPE Regressor achieves zero prediction error across MAE, MSE, and RMSE, with an R2 score of 1.0, demonstrating highly accurate NDVI prediction. These results confirm the effectiveness of the hybrid recurrent and ensemble framework for modelling modeling vegetation patterns.

Table. Table 2: Regression Results – NDVI Index Prediction Model Name MAE MSE RMSE R2 Score SVM 0.0836 0.0079 0.0890 0.7301 AB 0.1359 0.0238 0.1543 0.1892 Ridge 0.1487 0.0289 0.1700 0.0159 HRPE 0.0000 0.0000 0.0000 1.0000 Table 3 shows regression performance for predicting the number of days remaining until harvest. Traditional models such as Support Vector, AB, and Ridge exhibit higher prediction errors due to limited capability in capturing complex seasonal agricultural patterns. Support Vector regression shows unstable performance with a negative R2 score, indicating poor model fitting. The Proposed HRPE Regressor achieves zero error across MAE, MSE, and RMSE metrics, with a perfect R2 score. The results demonstrate the ability of the hybrid model to accurately estimate crop maturity timelines and support intelligent harvest planning.

Table. Table 3: Regression Results – Days to Harvest Prediction Model Name MAE MSE RMSE R2 Score SVM 83.5412 13676.1289 116.9450 -0.0465 AB 84.6583 9849.1149 99.2427 0.2463 Ridge 89.4087 12426.5254 111.4743 0.0491 HRPE 0.0000 0.0000 0.0000 1.0000 5. Conclusion The study introduces presents an advanced intelligent agricultural analysis framework that combines integrates one classification task and two regression tasks within a single integrated unified learning environment. The system is designed was developed to efficiently process analyse agricultural data efficiently and generate predictions for crop disease status, NDVI index, and days to harvest using SVM-1CA2RT, AB-1CA2RT, R-1CA2RT, and the proposed HRPE-1CA2RT model. Experimental findings indicate results show that the proposed HRPE-1CA2RT model outperforms all other models, achieving achieved the best overall performance, delivering 100% accuracy, precision, recall, and F1-score for crop disease classification, along with an R2 score of 1.0000 for both NDVI and harvest prediction. When compared to with SVM, AB, AdaBoost, and Ridge models, the proposed hybrid approach demonstrates significant demonstrated major performance gains improvements by effectively learning capturing complex nonlinear relationships patterns more effectively and producing more consistent predictions. highly reliable outputs. The combination integration of the RPN Recurrent Polynomial Network with ETT models enhances Ensemble Tao Tree methods significantly enhanced prediction accuracy, minimizes error, quality, reduced error values, and improves improved overall system robustness. Additionally, adaptability. In addition to strong predictive performance, the framework incorporates also provides secure user management, interaction, visualization capabilities, support, and efficient model storage, handling, making it highly suitable for real-world practical agricultural applications. analysis.