`Open Source AI Model Accurately Predicted Olive Oil Prices - Olive Oil Times

Open Source AI Model Accurately Predicted Olive Oil Prices

By Ofeoritse Daibo
Aug. 7, 2024 15:16 UTC

Olive oil prices around the world have reached his­toric highs due to poor har­vests caused by drought and high spring­time tem­per­a­tures in key pro­duc­ing coun­tries, most notably Spain.

While prices at ori­gin have fallen from their mid-January record high, retail prices remain well above aver­age, lead­ing to changes in shop­per behav­ior in tra­di­tion­ally heavy olive oil con­sumers.

Leading experts in the sec­tor antic­i­pate prices to decline based on expec­ta­tions of global olive oil pro­duc­tion return­ing to about three mil­lion tons in the 2024/25 crop year. However, the future direc­tion of prices remains far from cer­tain.

See Also:Researchers Introduce AI Tool to Help Olive Farmers Predict Harvest Timing

To that end, a researcher in the south­ern Spanish autonomous com­mu­nity of Andalusia, the world’s largest olive oil-pro­duc­ing region, devel­oped an arti­fi­cial intel­li­gence (AI) pro­gram to pre­dict olive oil prices.

Diego Hueltes, a com­puter engi­neer and the chief exec­u­tive of TADIA.ai, an AI con­sul­tancy, devel­oped three pre­dic­tive price mod­els, includ­ing a one-week fore­cast, a four-week fore­cast and a pre­dic­tion of whether prices will go up or down.

I come from Alcalá la Real, a town in Jaén where olive farm­ing is the main liveli­hood, and there’s a deep-rooted cul­ture around olive oil,” he told Olive Oil Times.

I found it fas­ci­nat­ing that the elders could pre­dict mar­ket move­ments based on weather and prices,” he added. If they could do it based on expe­ri­ence, my premise was that an AI model could achieve the same in a more sys­tem­atic way.”

Hueltes used auto­matic learn­ing algo­rithms and arti­fi­cial intel­li­gence research to ana­lyze his­toric olive oil prices, mete­o­ro­log­i­cal data and pro­duc­tion data to make accu­rate price esti­mates in 2017 and 2018.

briefs-business-europe-open-source-ai-model-accurately-predicted-olive-oil-prices-olive-oil-times

Diego Hueltes

I devel­oped two mod­els,” Hueltes said. The first one could pre­dict the price, an absolute num­ber. In this case, my model had a mean error of three per­cent. For exam­ple, if my model pre­dicts that next week the price will be €10.00, the actual price will prob­a­bly be between €9.97 and €10.03.”

The sec­ond model pre­dicted next week the price will go up’ or next week the price will go down,’ mak­ing it very straight­for­ward,” he added. It accu­rately pre­dicted the price direc­tion 76 per­cent of the time.”

To eval­u­ate this model, I used a finan­cial tech­nique that ana­lyzes returns by see­ing what the accu­mu­lated return would be if, each time the model pre­dicted the price would rise next week, I bought olive oil, and each time it pre­dicted the price would fall, I sold olive oil,” Hueltes con­tin­ued. If I had invested €100 fol­low­ing the model between 2017 and 2018, I would have €140 after a year.”

However, many things have hap­pened since 2018, includ­ing United States tar­iffs on Spanish olive oil, the Covid-19 pan­demic, the Russian inva­sion of Ukraine and Europe’s his­toric drought.

The model would­n’t be valid for 2024 even though its sci­en­tific basis remains valid,” said Hueltes, who pub­lished his research on GitHub, mak­ing it avail­able to the gen­eral pub­lic.

I’m a huge fan of open source, and I believe sci­en­tific research, espe­cially in AI, should be uni­ver­sally acces­si­ble and free,” he said. Since there was very lit­tle open and acces­si­ble work in the field of olive oil, I decided to make all my research avail­able.”

Hueltes hopes that this tech­nol­ogy will pos­i­tively impact the world’s largest olive oil-pro­duc­ing province in the future.

The AI model, using price, pro­duc­tion and mete­o­ro­log­i­cal data, can extract the under­ly­ing pat­terns in the data,” he said. These pat­terns, which are very dif­fi­cult for humans to detect, even for experts in the sec­tor, are entirely fea­si­ble for AI mod­els.”

Advertisement

According to Hueltes, pre­dict­ing olive oil prices allows for the opti­miza­tion of the oil trade, espe­cially for small pro­duc­ers, as they receive indi­ca­tors on whether prices are fair or not, or whether it is bet­ter to wait or act quickly.”

Additionally, the research could ben­e­fit actors across the olive oil value chain, includ­ing farm­ers, retail­ers and dis­trib­u­tors, con­sumers, investors and traders.

Farmers and pro­duc­ers can more effec­tively plan their sales strate­gies to max­i­mize prof­its and man­age resources effi­ciently.

Retailers and dis­trib­u­tors also can use accu­rate price pre­dic­tions to man­age inven­tory and develop pric­ing strate­gies that remain com­pet­i­tive and prof­itable.

Meanwhile, con­sumers can make informed pur­chas­ing deci­sions based on poten­tial price changes, such as buy­ing in bulk when prices are expected to rise.

Finally, investors and com­mod­ity traders can make bet­ter invest­ment deci­sions and increase their returns. Understanding price trends can also help for­mu­late poli­cies to sup­port the agri­cul­tural sec­tor and sta­bi­lize the mar­ket.

While Hueltes said the research may not directly impact cli­mate change, the symp­toms of which are widely attrib­uted to the recent pro­duc­tion declines, it can help under­stand how cli­mate fac­tors affect olive oil pro­duc­tion.”

The open-source nature of the research allows for the devel­op­ment of new mod­els to make the olive oil trade more equi­table,” he added.


Advertisement
Advertisement

Related Articles