Born in India I have seen many cases of TB. Tuberculosis (TB) is an airborne infectious disease that is preventable and curable but in some countries there is a stigma and considerable discrimination towards the TB sufferers. Data Geek Challenge was the motivator for me to personally analyse and share the trends of this deadly disease. I was able to analyse and predict the trends in about 45mins (of which 20mins was downloading the software )

WHO is working to dramatically reduce the burden of TB, and halve TB deaths and prevalence by 2015. Data provided by countries to WHO andestimates of TB burden generated by WHO for Global Tuberculosis Report 2012 was used to analyse the trends. For more info please refer too –

SAP Predictive Analysis was used to forecast the number of TB death in India by using Triple exponential smoothing which takes into account seasonal changes as well as trends (below are the algorithm variables used) –

Triple Exponential Smoothing Summary:

StartYear : 1995

Alpha : 0.5 / Beta : 0.8/ Gamma : 0.5

The Base : 60241.94353238099

The Trend : 242.7169517519365

The Seasonal Indices : 1.3062963825520746

R-square factor : 1.1747219022589726

Mean Squared Error : 2.1647103181288164E7

Root Mean Squared Error : 4652.644751245056

Mean Absolute Percent Error : 14.013437896853375

Goodness of fit : 0.9730434333921768

fvalue : 697.2531298065552

Top 5 countries – Number of TB deaths in the world.


TB Deaths – Actual vs Predicted Forecast for India


(Please note: All figures shown are for illustration purposes only)

To report this post you need to login first.


You must be Logged on to comment or reply to a post.

    1. Nameeth Raman Post author

      Hi Lee,

      None, the colour highlighted were manually changed (right-click
      on the label). Btw, you make a good point about predicting geospatial spread of
      an infection – will try.



  1. Boris Bader

    Hi Raman,

    nice analysis. When you take a close look to the last 4 years of infections it seems like the results are stagnating or even decreasing. Maybe you can get a even better solution with a higher alpha and beta factor. So new values will take more account.



    1. Nameeth Raman Post author

      Hi Boris,

      Thanks. You are right, I tried with higher α (data) and β (trend) smoothing factor and the

      prediction for last four years was leaning much closer to the dropping actual value – which is good!





Leave a Reply