Predictive analysis of De Tijdloze with the ARIMA model

Wanneer:

Tuesday, 11 february 2020,  16:00 – 16:30

Subject:

Predictive analysis of De Tijdloze with the ARIMA model

Topics:

• History of our ‘De Tijdloze’ Power BI report
• Analysis of different predictive models (RapidMiner, Azure Machine Learning Studio, Python …)
• Python script in the AutoRegressive Integrated Moving Average (ARIMA) model
• The results

Webinar hosted by

Wout Cardoen, Data Wizard at Kohera NV

Voor:

• Data Scientist
• Data Engineers
• BI consultants
• …

Voorspellen met Python in het ARIMA-model

Wout Cardoen, data wizard at Kohera

At the end of 2019 we traditionally provided an update to our Power BI report. This year we were slightly more ambitious and we attempted to predict the full Tijdloze 2019, based on a predictive analysis. Various technologies were used in our search for a suitable model. Ultimately, the model with time series analysis turned out to be the best predictive model. The AutoRegressive Integrated Moving Average (ARIMA) model is a static method for making predictions based on a timeline (in our case the edition of De Tijdloze).

The result was impressive: At position 93, we saw our first match and the closer we got to the top 10, the better our prediction. We had predicted nine of the ten numbers from the top 10 and three of them also had the predicted ranking.