Question

我为每位客户提供以下月度数据。

我想为每个客户做3个月的预测。

注意：很多obs都有零（没有事务）------需要处理这个稀疏数据集

         CustomerName        01/2009 02/2009 03/2009 04/2009 05/2009 06/2009 07/2009 08/2009 09/2009 10/2009
         Aaron Bergman         0.00    0.00    0.00    0.00    0.00     0.0 4270.87    0.00    0.00       0
         Aaron Hawkins         0.00    0.00    0.00    0.00    0.00     0.0    0.00  455.04    0.00       0
         Aaron Smayling        136.29  4658.69 0.00    119.34  4674.16  0.0    0.00    0.00    0.00       0
         Adam Bellavance       0.00    0.00    0.00    0.00    2107.55  0.0    0.00    0.00    0.00       0
         Adam Hart             60.52   0.00    0.00    0.00    0.00     0.0    0.00    0.00    0.00       0
         Adam Shillingsburg    0.00    1749.50 125.86  0.00    0.00     5689.4 3275.74 1296.30 9887.52    0
         Adrian Barton         0.00    66.00    0.00    0.00    0.00     55.0    0.00    0.00    0.00     0
         Adrian Hane           0.00    23.66    0.00    0.00    46.22     0.0    0.00    0.00    0.00     0
         Adrian Shami          10.00    0.00    0.00    33.00    0.00    48.0    0.00    0.00    42.00    0
         Aimee Bixby           56.33    22.99    0.00    44.28    0.00     0.0    0.00   66.12   0.00    48.22

我如何进行某种批次系列预测，比如为每个客户使用auto.arima ......

Answer 1

假设上述数据被称为df

(df_melt<-reshape2::melt(df))

# Sort by vector `CustomerName` then `variable`
df_melt_order<-df_melt[with(df_melt, order(CustomerName, variable)),]
df_melt_order<-df_melt_order[,c(2,1,3)] # reordering the columns
head(df_melt_order)

library(dplyr)
(model_fits <- group_by(df_melt_order, CustomerName) %>% do(fit=auto.arima(.$value)))
#The result is a data frame containing the model fits for each CustomerName:

#You can get a list with each model fit like so:
(fitted<-model_fits$fit)
sapply(fitted, function(x) plot(forecast(x, 10))) # forecast plot for each time series

仍然无法从上面的模型拟合得到点预测......需要一些帮助

如何进行多个时间序列预测，即为每个客户进行预测

1 个答案: