Leveraging feature engineering to boost placement predictability for seed product selection and recommendation by field

1 . A computer-implemented method comprising:

receiving, by a server computer system, a plurality of agricultural data records, which include a yield property(ies) of products grown in a given field and continuous data indicative of multiple raw field features specific to the given field;

identifying ones of the agricultural data records, for which the products grown in the given field include multiple products concurrently grown in the given field;

identifying, in the identified ones of the agricultural records, a subset including multiple ones of the multiple raw field features of the products, the multiple ones of the multiple raw field features including at least one soil feature and/or at least one topography feature;

for each raw field feature included in the subset:

transforming, by the server computer system, the continuous data for the raw field feature into one of multiple distinct feature classes based on a value of the continuous data of the raw field feature relative to distinct numeric classification criteria of the feature classes;

generating, by the server computer system executing a best linear unbiased prediction model, based on the distinct feature classes but not the continuous data for the subset of the multiple ones of the multiple raw field features, genomic-by-environmental relationships between one or more of the products grown in the field;

generating, by the server computer system executing a statistical mixed effects model, based at least in part on the genomic-by-environmental relationships between the one or more of the products grown in the field, predicted yield performance for a set of products associated with one or more target environments thereby capturing variability in yield for the set of products under different environmental conditions for the one or more target environments;

generating, by the server computer system executing a recommendation model, product recommendations for the one or more target environments based on the predicted yield performance for the set of products; and

providing one or more instructions configured to cause display, on a display device communicatively coupled to the server computer system, of the product recommendations.

2 . The computer-implemented method of claim 1 , wherein generating the genomic-by-environmental relationships includes generating the genomic-by-environmental relationships between genetic features of the one or more products, the transformed multiple raw field features of the given field, and the yield property (ies) of the one or more products.

3 . The computer-implemented method of claim 1 , wherein the identified subset further includes one or more of soil texture, soil drainage, crop rotation, tillage, field elevation, and/or field slope.

4 . The computer-implemented method of claim 1 , wherein the product recommendations are for soybean varieties.

5 . The computer-implemented method of claim 1 , wherein the subset includes pH, soil cation-exchange capacity (CEC), and organic matter (OM); and

wherein the distinct numeric classification criteria for the feature classes include:

for the value of the pH, a high pH range greater than 7.0, a medium pH range between 5.8 and 7.0, and a low pH range less than 5.8; and

for the value of the soil CEC, a high CEC range greater than 20, a medium CEC range between 10 and 20, and a low CEC range less than 10.

6 . The computer-implemented method of claim 1 , further comprising:

generating, by the server computer system, based on the product recommendations, one or more scripts;

transmitting, by the server computer system, the one or more scripts to an application controller of an agricultural vehicle; and

executing the one or more scripts to automatically control an operating parameter of the agricultural vehicle to plant seeds in the one or more target environments consistent with the product recommendations.

7 . One or more non-transitory computer-readable storage media storing instructions which when executed by one or more processors cause performing operations comprising:

receiving a plurality of agricultural data records, which include a yield property (ies) of one or more products grown in a given field and continuous data indicative of multiple raw field features and specific to the given field;

identifying ones of the agricultural data records, in which the products grown in the given field include multiple products concurrently grown in the given field;

identifying, in the identified ones of the agricultural records, a subset including multiple ones of the multiple raw field features of the products, the multiple ones of the multiple raw field features including at least one soil feature and/or at least one topography feature;

for each raw field feature included in the subset:

transforming the continuous data for the raw field feature into one of multiple distinct feature classes based on a value of the continuous data of the raw field feature relative to distinct numeric classification criteria of the feature classes;

generating, using a best linear unbiased prediction model, based on the distinct feature classes, but not the continuous data for the subset of the multiple ones of the multiple raw field features, genomic-by-environmental relationships between one or more of the products grown in the field;

generating, using a statistical mixed effects model, based at least in part on the genomic-by-environmental relationships between the one or more of the products grown in the field, predicted yield performance for a set of products associated with one or more target environments thereby capturing variability in yield for the set of products under different environmental conditions for the one or more target environments;

generating, using a recommendation model, product recommendations for the one or more target environments based on the predicted yield performance for the set of products; and

providing one or more instructions configured to cause display, on a display device communicatively coupled to the one or more processors, of the product recommendations.

8 . The one or more non-transitory computer-readable storage media of claim 7 , wherein the operation of generating the genomic-by-environmental relationships includes generating the genomic-by-environmental relationships between genetic features of the one or more products, the transformed multiple raw field features of the given field, and the yield property (ies) of the one or more products.

9 . The one or more non-transitory computer-readable storage media of claim 7 , wherein the product recommendations are for soybean varieties.

10 . The one or more non-transitory computer-readable storage media of claim 7 , wherein the identified subset further includes one or more of soil texture, soil drainage, crop rotation, tillage, field elevation, and/or field slope.

11 . The one or more non-transitory computer-readable storage media of claim 7 , wherein the subset includes pH, soil cation-exchange capacity (CEC), and organic matter (OM); and

wherein the distinct numeric classification criteria for the feature classes include:

for the value of the pH, a high pH range greater than 7.0, a medium pH range between 5.8 and 7.0, and a low pH range less than 5.8; and

for the value of the soil CEC, a high CEC range greater than 20, a medium CEC range between 10 and 20, and a low CEC range less than 10.