Data science shows it can-can by predicting Strictly outcome

Each week we rank Strictly contestants based on social media analysis. We call this the Strictly Sentiment score.

Week two and this time it's for real. It's the first elimination.

We analysed thousands of tweets about each contestant to extract public sentiment - are people positive, negative or neutral about the contestant when tweeting? We then ranked them on the basis of this Strictly Sentiment. 

Maisie and HRVY topped the judges leaderboard this week but who is top of the Sentiment poll?

This week Bill blows everyone out of the water! He has the most positive tweets by a long distance. Nicola, Maisie and HRVY are next in line but still a long way behind Bill.

At the other end of the leaderboard JJ and Jamie struggle for sentiment.

Remember, the Strictly Sentiment leaderboard is based upon Tweets. These may or may not reflect the voting public's choices.



Strictly Sentiment leaderboard

We've combined the judges scores with the Strictly Sentiment result to show how the final rankings would have looked if social media rather than a public vote was used. As with the show itself we have assigned scores based on where each contestant ranks.

The judges scores for weeks 1 and 2 have been combined and we've done similar with the sentiment score. Based on a combination of Strictly Sentiment and judges scores Jacqui and Jamie fell into dance off peril.

The outcome was the same when the public vote was taken into account. Jacqui and Jamie danced off with the judges voting to save Jamie and eliminate Jacqui.

So how did the sentiment analysis do? The BBC do not release the results of the public vote so we are not able to do a comparison. All we can compare against is the outcome from combining judges score and sentiment scores - on this metric, the model predicted a successful outcome.

If, however, the elimination had been based purely on the Sentiment score then Jacqui would have been safe and JJ and Jamie would have been in the bottom two. Does this mean that social media was more positive towards Jacqui than the voting public? This is quite possible but it is also fair to say that Jacqui was undone by a very low judges score which made it very difficult for her to escape the dance off without a solid public vote behind her.

We will do it all again next week. Don't forget that we tweet our prediction one hour before the results show on Sunday. Follow us on Twitter if you'd like to know the prediction.

Strictly Sentiment score

The Strictly Sentiment score is derived from a sentiment analysis of tweets. Using natural language programming and machine learning we classify each tweet as positive, neutral or negative. Tweets are filtered so that they only relate to a celebrity's appearance on the show rather than what they do in their 'day job'. Only tweets made between 7pm and midnight are included. This is to ensure that tweets relate to Strictly.

We then add up all the positive tweets for each contestant and assign a score between 1 and 100.  This score is based on the relative distribution of positive tweets. We do this to make it easier to compare and contrast Strictly Sentiment scores.