Project Summary

Is the sentiment of twitter data about specific car brands, related to the stock market returns of the brand?

This repository contains all of the code required to set-up the twitter-listener, collect data and construct a machine-learning model that analyses the daily sentiment of tweets related to automotive companies. Finally, once the time series of the daily sentiments are calculated using a Bayesian-weighted average, a Bayesian linear regression model is fitted relating the calculated sentiment time-series aswell as the daily market return to the daily return of the automotive stock.

Using the Bayesian linear regression, it is then possible to generate posteriors for the model parameters with qunaitfied uncertainties, and finally place bounds on the correlations between twitter sentiments and the daily returns of the stock.

Sentiment Analysis of Automotive Companies

This repository carries out sentiment analysis on website review data from different sources. Please look at the main Jupyter notebook

This code was written for a sentiment analysis model for ApiThinking.

Setup instructions

1. Install conda

Choose the version of Anaconda that you wish to download, for example

$  wget https://repo.continuum.io/archive/Anaconda3-2018.12-Linux-x86_64.sh  
$ bash Anaconda3-2018.12-Linux-x86_64.sh

2. Create Conda enviroment

After conda has been installed. The enviroment to run the code needs to be created. The enviroment that is used for this program is found in py35.yml. To create this enviroment:

$ conda env create -f py35.yml

3. Run conda enviroment

To run the conda envioment that was created,

$ source activate py35

It may be necessary to download Spacys word embedding model,

python -m spacy download en_core_web_lg

4. Export changes to conda enviroment

If you have made changes to the enviroment needed to run the code in the repository, export the enviroment using,

$conda env export > <environment-name>.yml

Repository contents

1. Interactive Sentiment analysis model

This folder contains all steps in the project from training the model using a combination of Amazon customer review data along with Yelp review data. The final chosen model was voting ensemble of a Naive Bayes Classifier, Logistic regression and the TextBlob prebuilt classifier.

2. Automated Sentiment Analysis code

This folder contains the code that will carry out all steps in the analysis automatically, according to the specified input parameters.

3. Twitter data collection code

This folder contains all of the code required to run the Twitter data listener and collect the twitter data as a SQL database.