FieldInsights
Product Specification
Product Name:Agronomic Modeling Training and Evaluation
Product SKU:71550-00
Revision Date:2024-01-08
Product Overview
Description
Agronomic Modeling Training and Evaluation includes a tailored consultation with Sentera's Agronomic Modeling Scientists to achieve the overall goal of training and cross-validating one or more customized agronomic model(s) that meet your specific objective(s). The primary steps in this process include (but are not necessarily limited to):
- i) exploratory analysis on available data (to identify the expected environmental and management constraints of the trained model)
- ii) feature engineering based on your hypotheses and/or available feature data
- iii) hyperparameter tuning
- iv) model experimentation, including training, cross-validation, and testing for the desired statistical algorithms and cross-validation strategies
- v) reporting/communication of the deliverables.
Recommended Use Cases
- All Agronomic Modeling Products
Deliverables Summary
- Performance statistics for each trained agronomic model.
- Descriptive statistics for response variables and candidate features.
Requirements
- "Agronomic Modeling Training and Evaluation Plan" completed prior to project start date
- "Agronomic Modeling Customer Data Contribution" provided prior to project start date
- All "Integrated Project Team and Program Management" members identified
- Requires FieldAgent Enterprise License (61006-##) or Enterprise Analytics Upload/Export Surcharge (71010-09)
Agronomic Modeling Training and Evaluation Plan
Prior to beginning the project (i.e. prior to any data sharing or model training), Customer will fill out an Agronomic Modeling Training and Evaluation Plan provided by Sentera.
The purpose of the Agronomic Modeling Training and Evaluation Plan is to gain an understanding of the following:
- Agronomic objectives (what is Customer trying to solve?)
- Historical data inventory (what types of data are available for addressing the agronomic modeling objectives?)
- Response variable(s)
- Feature data
- Number of observations
- Number of years, field trials, locations, etc. represented
- All shared data observations must include explicit spatial and temporal information (i.e., where and when was it collected/measured?)
- Data sharing plan (for Customer proprietary data)
- Additional requirements may be identified based on the exact nature of agronomic modeling objectives and/or historical data inventory
Agronomic Modeling Customer Data Contribution
All customer data identified as part of the Agronomic Modeling Training and Evaluation Plan will be provided prior to beginning the project.
Integrated Project Team and Program Management
Sentera is experienced in delivering complex projects for its Customers using best practices in project management and execution, including the formation of a dedicated Integrated Project Team to ensure project success through regular status and update meetings.
Sentera proposes an Integrated Project Team to execute this project with team member assignments including:
- Sentera Project Manager
- Customer Project Lead
- Customer Field Supervisors or Technicians (point of contact for each site)
The Sentera project manager will provide overall management and coordination for Sentera's effort on this project. The Sentera project manager will work directly with the customer's assigned project lead(s).
After an initial kick-off meeting, short weekly or bi-weekly status calls will be held by the integrated project team leading up to and during key phases of the project. This ensures project success through scheduling data collections to match conditions on the ground and provide useful feedback throughout the course of the project. Project schedules will be updated throughout the project duration.
Deliverables
Analytics
| Analytic |
|---|
| Model Training and Evaluation (89001-00) |
Model Training and Evaluation (89001-00)
Description
Model Training and Evaluation provides both performance statistics for trained agronomic model and descriptive statistics for response variables and features.Deliverable List
| Deliverable | Format | Description |
|---|---|---|
| Trained Agronomic Model Performance Statistics | CSV | Goodness-of-fit statistics for each trained agronomic model |
| Trained Agronomic Model Performance Visualization | PNG | Visualization of 1:1 line with measured vs. predicted values for each trained agronomic model |
| Response Variable and Feature Descriptive Statistics | CSV | Descriptive statistics for each response variable and feature |
Data Attributes
Trained Agronomic Model Performance Statistics
| Attribute | Description | Units Metric | Units Imperial |
|---|---|---|---|
| model_name | Name of trained agronomic model | - | - |
| model_type | Algorithm used for training agronomic model | - | - |
| n_features | Number of features used in the trained agronomic model | - | - |
| feature_list | Ranking of features by importance for trained agronomic model | - | - |
| response | Name of response variable | - | - |
| n | Total number of observations used | - | - |
| cv_strategy | Cross validation strategy used for training agronomic model | - | - |
| n_train | Number of observations in the "train" split | - | - |
| n_test | Number of observations in the "test" split | - | - |
| r2_train | Coefficient of Determination [R2] for the "train" split | – | – |
| r2_test | Coefficient of Determination [R2] for the "test" split | – | – |
| rmse_train | Root Mean Squared Error [RMSE] for the "train" split | – | – |
| rmse_test | Root Mean Squared Error [RMSE] for the "test" split | – | – |
| rmse_rel_test | Relative Root Mean Squared Error [RMSE] for the "test" split | % | % |
| mae_train | Mean Absolute Error [MAE] for the "train" split | – | – |
| mae_test | Mean Absolute Error [MAE] for the "test" split | – | – |
| mae_rel_test | Relative Mean Absolute Error [MAE] for the "test" split | % | % |
Response Variable and Feature Descriptive Statistics
| Attribute | Description | Units Metric | Units Imperial |
|---|---|---|---|
| variable | Name of response variable or feature | - | - |
| mean | Mean value of response variable or feature | – | – |
| median | Median value of response variable or feature | – | – |
| std_dev | Standard deviation of response variable or feature | – | – |
| min | Minimum value of response variable or feature | – | – |
| max | Maximum value of response variable or feature | – | – |