Data to rescue? Volume prediction, production control, and food security

Data to rescue? Volume prediction, production control, and food security

In Aug 2019, the Project Transition team visited a few villages in Telangana. These villages were unlike the rest. They had an impressive number of people doing organic and natural farming. They, unlike many farming communities across the country, believed that this mode of farming is a pathway for a sustainable financial future. Their optimism is not unfounded. They usually sell out all that they produce, and routinely fall short of meeting the actual market demand. This essentially translates to little or no financial loss for them. This also has the effect of keeping their produce at a premium in the market. 

This raises an interesting point, pertinent to enhancing food security; which is a larger question of larger import to the society. Food security correlates strongly to the food production potential and loss-free food flow from “farm to fork”. If done right, it would also enhance the incomes of the farmer. 

The question then is: would it be possible to offer forecasting to the farmers, through data science tools, so that they make the best of market potential? A positive financial impact enabled via this route, could ameliorate the desperate measures we see in farming practice today. Arguably, chemical farming can be labelled a desperate measure, and a means of reducing uncertainties. At least some of these uncertainties could be abated(?), leading to better options for both the farmer and the consumer. 

We wouldn’t be able to answer the above questions comprehensively without focused research efforts. However, we know that forecasting algorithms exist. Large data and machine learning based predictions have been game changers elsewhere within the economy (financial engineering and e-shopping industry being the most notable examples). We will avoid the details here. The idea is to place such robust forecasting based paradigms, and aid decision making in farming and food supply chains. 

What would such a paradigm entail? Well…it would involve volumes data (supply, demand, loss) from across all points in the supply chain, in a time series manner. Then carefully chosen regression and statistical equation models would need to be used to deliver predictions, with an estimated error bar. Recommender systems would convert these into “suggestions” understandable to the stakeholders of the supply chain. 

This seems eminently doable – is it not? But how would a firm that offers such technical services be a profitable and financially sustainable agency? I believe it will be through a subscription model wherein each node in the supply chain could pay a small fee for data supplied, and the notional certainty offered. Regulations would be needed eventually to ensure reasonable tariffs for the services. I see enterprise possibilities here for young technopreneurs, and new vistas for private-public partnerships. 

To summarize, a volume prediction and production-recommendation app would be sweet. This is doable and could be built ground up! Teams like Transition can assist in achieving such goals. 

– by Dr. Tiju Thomas

Department of Metallurgical & Materials Engineering, IIT Madras