我尝试在以下数据中使用LIBSVM回归预测6个月： 我会使用带有RBF内核的LIBSVM和带有默认数据的SVMTType-SVR（我不是专家来修改它）
由于我在训练数据上评估的数据很少，但我每6个月收到0.5804的预测，所以看来出现了问题？我试图在weka中使用网格搜索来寻找优化参数，但结果是一个不变的预测。
有没有人有一些改善我的建议？
谢谢
@relation Regression_test
@attribute Period date yyyy-MM-dd
@attribute Percentage numeric
@data
2011-08-01,0
2011-09-01,0
2011-10-01,0.259403
2011-11-01,0.308642
2011-12-01,0.613497
2012-01-01,2.037662
2012-02-01,1.134486
2012-03-01,0.898727
2012-04-01,0.465357
2012-05-01,0.241168
2012-06-01,0.354359
2012-07-01,0.468987
2012-08-01,0.320699
2012-09-01,0.584155
2012-10-01,0.786552
2012-11-01,0.385395
2012-12-01,0.302407
2013-01-01,0.490101
2013-02-01,0.446799
2013-03-01,0.328194
2013-04-01,0.431381
2013-05-01,0.445664
2013-06-01,0.557984
2013-07-01,0.735813
2013-08-01,0.82297
2013-09-01,0.838937
2013-10-01,1.06926
2013-11-01,0.773331
2013-12-01,0.408285
2014-01-01,0.271486
2014-02-01,0.394293
2014-03-01,0.394572
2014-04-01,0.563725
2014-05-01,0.878016
2014-06-01,0.766179