This is project I did as part of my Data-Scientist Nanodegree with udacity.It is a disaster response pipeline which I bulit on the figure eight data-set. It helps disaster recovery team to quickly see which requests are emergency by using text classification. It includes flask and machine learning at back-end.
This is an standard ETL pipeline where data is preprocessed,cleaned,and RAW data is getting transformed to new data which can be used by the machine learning models.
In this pipeline we have used sklearn pipeline to bulit ml pipeline with countvectorizer,tfidftransformer,verb extractor and one estimator(Multioutputclassifier) Adaboostclassifier.
(Note -: I tried using random forest but found out by trial and error that Adaboostclassifier has better evaluation score)
Step 1-: Clone the project to local system.
Step 2 -: Go to the Disaster-Response-Pipeline/app folder using 'cd' command in cmd or terminal.
Step 3 -: Run 'python run.py'.
Step 4 -: open 'http://localhost:3001/' in browser.
step 5 -: HAVE FUN
-
app-: This folder has two file one templates and one app.py. app.py is the main file from which the project is running.
-
data -This folder has two csv files which are the files provided by figure eight company for this project,their is db file which gets created when ETL code in the folder gets run.
-
model-This folder has one ml pipeline code and one pickle file.Pickle file is the trained file we get after ml code is run in the folder.
Mohit Jawale