Agricultural data integration and analysis platform
Provided herein are methods, systems, and media that implement machine learning algorithms to determine a cultivar regimen recommendation for a crop based on crop yield and cultivar condition data.
1 . A computer-implemented method of training a neural network for determining a cultivar regimen recommendation, the method comprising:
a) collecting from a database a plurality of historical growing conditions, wherein each historical growing condition comprises a historical cultivar condition and a historical cultivar regimen, and wherein the historical growing condition is associated with a historical crop yield;
b) creating a first training set comprising:
i) a first plurality the collected historical growing conditions wherein each historical growing condition is associated with the historical crop yield; and
ii) a second plurality the collected historical growing conditions wherein each historical growing condition is disassociated from the historical crop yield;
c) training the neural network in a first stage using the first training set to determine a predicted crop yield;
d) creating a second training set for a second stage of training comprising the first training set and one or more of the second plurality of the collected historical growing conditions wherein a difference between the determined crop yield and the predicted crop yield is greater than a set amount; and
e) training the neural network in a second stage using the second training set;
wherein the cultivar regimen recommendation, the historical cultivar regimen, or both comprise a fertilizer quantity adjustment, a fertilizer adjustment, a pruning quantity adjustment, a pruning location adjustment, a pesticide quantity adjustment, a pesticide adjustment, a planting date adjustment, a harvesting date adjustment, an irrigation quantity adjustment, an irrigation time of day adjustment, an irrigation schedule adjustment, a crop adjustment, or any combination thereof; and
wherein the historical growing condition comprises a wind speed, a wind direction, a gust speed, a gust direction, a rainfall quantity, a soil moisture, a light measurement, a humidity, a crop dimension, a soil pH, subsurface plant and/or tree root growth and propagation, above surface plant and/or tree growth and propagation, soil textural properties, soil temperature, soil ion concentration, soil pore fluid composition, a gamma-ray measurement, a picture, an audio track, a video, aerial imagery, satellite imagery, a chemical composition, an atmospheric pressure, an O 2 quantity, a N 2 quantity, a CO 2 quantity, a sporadic light measurement, a fruit growth measurement, a reflectance, an infrared measurement, a mid-infrared measurement, near-infrared measurement, a fruit density, a GPS position, a temperature, or any combination thereof.
2 . A computer-implemented system comprising: a digital processing device comprising: at least one processor, an operating system configured to perform executable instructions, a memory, and a computer program including instructions executable by the digital processing device to create an application to train a neural network to determine a cultivar regimen recommendation, the application configured to perform at least the following:
a) collecting from a database a plurality of historical growing conditions, wherein each historical growing condition comprises a historical cultivar condition and a historical cultivar regimen, and wherein the historical growing condition is associated with a historical crop yield;
b) creating a first training set comprising:
i) a first plurality the collected historical growing conditions wherein each historical growing condition is associated with the historical crop yield; and
ii) a second plurality the collected historical growing conditions wherein each historical growing condition is disassociated from the historical crop yield;
c) training the neural network in a first stage using the first training set to determine a predicted crop yield;
d) creating a second training set for a second stage of training comprising the first training set and one or more of the second plurality of the collected historical growing conditions wherein a difference between the determined crop yield and the predicted crop yield is greater than a set amount; and
e) training the neural network in a second stage using the second training set;
wherein the cultivar regimen recommendation, the historical cultivar regimen, or both comprise a fertilizer quantity adjustment, a fertilizer adjustment, a pruning quantity adjustment, a pruning location adjustment, a pesticide quantity adjustment, a pesticide adjustment, a planting date adjustment, a harvesting date adjustment, an irrigation quantity adjustment, an irrigation time of day adjustment, an irrigation schedule adjustment, a crop adjustment, or any combination thereof; and
wherein the historical growing condition comprises a wind speed, a wind direction, a gust speed, a gust direction, a rainfall quantity, a soil moisture, a light measurement, a humidity, a crop dimension, a soil pH, subsurface plant and/or tree root growth and propagation, above surface plant and/or tree growth and propagation, soil textural properties, soil temperature, soil ion concentration, soil pore fluid composition, a gamma-ray measurement, a picture, an audio track, a video, aerial imagery, satellite imagery, a chemical composition, an atmospheric pressure, an O 2 quantity, a N 2 quantity, a CO 2 quantity, a sporadic light measurement, a fruit growth measurement, a reflectance, an infrared measurement, a mid-infrared measurement, near-infrared measurement, a fruit density, a GPS position, a temperature, or any combination thereof.
3 . A non-transitory computer-readable storage media encoded with a computer program including instructions executable by a processor to create an application to train a neural network to determine a cultivar regimen recommendation, the application configured to perform at least the following:
a) collecting from a database a plurality of historical growing conditions, wherein each historical growing condition comprises a historical cultivar condition and a historical cultivar regimen, and wherein the historical growing condition is associated with a historical crop yield;
b) creating a first training set comprising:
i) a first plurality the collected historical growing conditions wherein each historical growing condition is associated with the historical crop yield; and
ii) a second plurality the collected historical growing conditions wherein each historical growing condition is disassociated from the historical crop yield;
c) training the neural network in a first stage using the first training set to determine a predicted crop yield;
d) creating a second training set for a second stage of training comprising the first training set and one or more of the second plurality of the collected historical growing conditions wherein a difference between the determined crop yield and the predicted crop yield is greater than a set amount; and
e) training the neural network in a second stage using the second training set;
wherein the cultivar regimen recommendation, the historical cultivar regimen, or both comprise a fertilizer quantity adjustment, a fertilizer adjustment, a pruning quantity adjustment, a pruning location adjustment, a pesticide quantity adjustment, a pesticide adjustment, a planting date adjustment, a harvesting date adjustment, an irrigation quantity adjustment, an irrigation time of day adjustment, an irrigation schedule adjustment, a crop adjustment, or any combination thereof; and
wherein the historical growing condition comprises a wind speed, a wind direction, a gust speed, a gust direction, a rainfall quantity, a soil moisture, a light measurement, a humidity, a crop dimension, a soil pH, subsurface plant and/or tree root growth and propagation, above surface plant and/or tree growth and propagation, soil textural properties, soil temperature, soil ion concentration, soil pore fluid composition, a gamma-ray measurement, a picture, an audio track, a video, aerial imagery, satellite imagery, a chemical composition, an atmospheric pressure, an O 2 quantity, a N 2 quantity, a CO 2 quantity, a sporadic light measurement, a fruit growth measurement, a reflectance, an infrared measurement, a mid-infrared measurement, near-infrared measurement, a fruit density, a GPS position, a temperature, or any combination thereof.