Time Series Analysis - Predict Next Number Algorithm

[ccondran08]

Hi, looking for some assistance here for a systematic way to help predict the next time series number after adding new numbers each day. I have 1,595 lines data that consists of the follow values from 0 to 4;
- 0 (694 times - 42.9%)
- 1 (649 times - 40.7%)
- 2 (226 times - 14.2%)
- 3 (31 times - 1.9%)
- 4 (5 times - 0.3%)
The data starts of with numbered on the left from 1 to 1595 then the numbers mentioned above in the middle and the accumulated sum on the right. This gives the accumulated sum column an equation of Y = 0.7608x + 3.6121 and R squared figure of 0.9998.
- 1 - 2 - 2
- 2 - 0 - 2
- 3 - 2 - 4
- . .
- 1594 - 2 - 1213
- 1595 - 1 - 1214
I am currently using a Nearest Neighbour algorithm to predict the number and can get around 40% predicted right using Microsoft Excel. If anyone has any information out there that can help me with this problem to increase my prediction rate then I will be greatly appreciated. I've been stuck on this formula and is doing my head in. Thanks in advance.

 

[oldbrewer]

if you had a million rows 1 could be the most frequent figure. You seem to be trying to guess short term blips. But anyway please tell us how the 0 to 4 values are produced, eg football scores, random number indicator, customer complaints per day. The type of data is a clue as to how to analyse it.