Predicting inflation

Inspiration

There is often uncertainty when making assumptions about inflation ahead of the ONS release of data and we didn't want to rely on traditional data to make predictions which often rely on past economic data.

What it does

It considers past data and the seasonality as well as sentimental values to gives you a prediction for inflation right now.

How we built it

We used an API by the Guardian news company, to get articles using keywords like "UK inflation" or "cost of living", which reflect sentiment of people in the UK, and by performing sentiment analysis using Venders we {insert waffle about model}

Challenges we ran into

We were rate limited when scraping data from the guardian in order to train the model which lead to sub-optimal results. There was also difficulties in communication when it came to working on the model, this is because of varying experience when it came to using git/github. There was also errors in translations as we all had a mix of degrees so concepts often did not translate well.

Accomplishments that we're proud of

We managed to make a model that could make inflation predictions with a 95% confidence level.

What we learned

We learnt how to use Git with a better mastery, understood different factors which could contribute to inflation as well as learning the different models when it came to predicting inflation. We also improved on communication skills such as working as a group and voicing opinions.

What's next for Predicting inflation

Implement a paid API to scrape larger dataset, and considering supply-side shocks using social media alongside news articles as well as considering Markov switching regime

Built With

• google-notebook
• guardian
• python