# Predict Values based on 16 values of Past 16 years

I have a dataset of total number of Vehicles Registered for 16 years. These are a total of 16 values from 2001-2016. Which Machine Learning Technique would be best for predicting the number of vehicles in the upcoming years - say till 2050 - while using R ?

This is my Dataset:-

         Date    Bikes    Cars
1  2001-01-01  2283381 1198918
2  2002-01-01  2341051 1279362
3  2003-01-01  2379260 1289854
4  2004-01-01  2609442  298353
5  2005-01-01  2649910 1318488
6  2006-01-01  2757842 1372191
7  2007-01-01  2895734 1440801
8  2008-01-01  3039815 1549854
9  2009-01-01  3215583 1657860
10 2010-01-01  4305121 1726347
11 2011-01-01  5781953 1881560
12 2012-01-01  7500182 2094289
13 2013-01-01  9064547 2281083
14 2014-01-01 10341326 2400690
15 2015-01-01 12177352 2531592
16 2016-01-01 12600402 2582149

• Is there a "1" missing in your number for cars in 2004? Feb 6, 2018 at 15:59
• Yes Sir. There is a "1" missing ! Feb 6, 2018 at 18:12

Look at .

Forecasting: Principles and Practice by Hyndman & Athanasopoulos is an excellent free online textbook which uses R and its forecast package throughout.

• Dear Stephan, thanks a lot for your answer. The link you provided does not appear to be working though ! Feb 6, 2018 at 17:37
• Both links work for me. Which link is giving you problems, and what kind of problems? Feb 6, 2018 at 18:52
• Sir, this Link which you shared in your answer above with the display text: Forecasting: Principles and Practice by Hyndman & Athanasopoulos. Feb 6, 2018 at 20:20
• The link works completely fine for me. I have recommended it many times on CV and never heard of any problems. Is it possible that it could be blocked for some strange reason on your end? Feb 7, 2018 at 7:50

With only 16 data points, machine learning techniques would not be a good way to approach your problem. What you likely want is some form of extrapolation. The key for the success of your prediction would be the "qualitative" story, or assumptions behind the extrapolation that you choose to use. You might want to think about why the numbers between 2001-2016 have been the way that you see, what drove them, and what might happen to these drivers in the future. Some of these drivers could be population growth and growth in prosperity. Forecasting either of those is not trivial, but you may be able to use forecasts made by other researchers.

You may also want to extrapolate the data into the future using different assumptions and show the projections under each assumption. Depending on the application, you may want to also assign a probability to each scenario.

• Thanks a lot Dear Rinspy. I am not able to upvote your answer but I do appreciate your response. Thanks once again. Feb 6, 2018 at 17:38