On the afternoon of March 30, 2024, the Agricultural Economics Forum and the 70th Anniversary of the Department/20th Anniversary of the School series lecture was held in Room 931 of the Mingde Main Building. Professor Yu Xiaohua from the Department of Agricultural Economics in Development and Transition Countries at the University of Göttingen, Germany, delivered an academic lecture titled “The ‘No Free Lunch Theorem’ and Algorithm Selection in Agricultural Policy Research: A Case Study of Using Machine Learning to Predict Changes in Pig Prices.” The lecture was hosted by Professor Mao Xuefeng, Vice Dean of the School, and attended by many faculty members and students from both within and outside the university, including Professors Zeng Yinchu, Tang Jianjun, Associate Professors Huang Bo, and Zhou Yang.

Firstly, Professor Yu Xiaohua pointed out that in the era of big data, the foundational theories of statistics need reconstruction, and traditional agricultural economics, which is impacted, must undergo transformation. He introduced the fundamental theory in machine learning known as the “No Free Lunch Theorem,” which states that, on average, no single model or algorithm is universally superior for all possible machine learning tasks. However, for specific problems, a particular algorithm might be superior under certain conditions. The practical applications of machine learning mainly include prediction, clustering recognition, reinforcement learning, and modal generation.

Professor Yu then compared econometric models and machine learning models, highlighting differences in model objectives, bias-variance trade-offs, model linearity assumptions, and model settings. He explained three criteria for evaluating the performance of machine learning models and used the example of "pig price prediction" to compare the use of four machine learning algorithms (ARIMA, CNN, RNN, LSTM) in predicting pig prices.

Finally, Professor Yu emphasized that traditional agricultural policy analysis focuses on analyzing the relationships between variables and estimating model parameters, whereas “machine learning” emphasizes the accuracy of predicting future outcomes, which aligns closely with the goals of agricultural policy analysis. He encouraged researchers to embrace machine learning with an open mind and to actively engage in cutting-edge interdisciplinary exploration.