Prediction of corn market price based on Data Mining
Abstract
The price fluctuations of agricultural products and its trend are closely related to
people's lives.The corn is one of the major grain production in China, and the
annual production is very large, and the planting is also very wide. What is more,
the consumption of the corn for food crops, feed crops, fuel crops and other aspects
is also great.The annual production and consumption of corn are large, and they are
closely related to the income of farmers, so in recent years, the grain yield, the
price and demand of corn are concerned.What is more, the price is one of the most
concerned aspect.
Over the years, many domestic and foreign scholars have done a lot of
researches, and many ways of the time series model are applied to various price
analysis and forecast.in the end, the effect is better than the linear model. Today,
the method of data mining is popular .There are many reasons. On the hand,the
ways of various algorithms has high accuracy and it has good effect , but also can
be applied to various fields, on the other hands, they are not limited by many
conditions. Also, the computing speed of various methods is quite fast. In recent
years, many scholars have applied data mining algorithms to the analysis of
agricultural products.
In general, we can use the way of the combination of statistics and data mining
algorithms solve practical problems quickly and well.I aims to solve practical
problems in this paper, and I choose the price of agricultural products - corn,which
will be researched.On the basis of ways of all the scholars,I consider and choose
the appropriate method .And I will select the optimal model for the final prediction
according to the actual problem.By using the R tools,I establish the time series
model, the neural network model to solve the problems. For time series model, the
AR (1), ARIMA (1,1,1) ARIMA (2,1,3), seasonal model and seasonal model
ARIMA (2,1,3)
and seasonal model ARMA(2,2)of the 5 models, are selected .For the neural
network model,I choose the first five months, six months and twelve months
respectively as input neurons,and choose future data of one month, two months,
three months, four months and five month as output neurons.And then,I set the
number of hidden units, back propagation rate and the number of iterations of the
