Analyzed disaster data from Figure Eight to build a model for an API that classifies disaster messages. This project includes a web app where an emergency worker can input a new message and get classification results in several categories. The web app also displays visualizations of the data.
-
Run the following commands in the project's root directory to set up your database and model.
- To run ETL pipeline that cleans data and stores in database
python data/process_data.py data/disaster_messages.csv data/disaster_categories.csv data/DisasterResponse.db - To run ML pipeline that trains classifier and saves
python models/train_classifier.py data/DisasterResponse.db models/classifier.pkl
- To run ETL pipeline that cleans data and stores in database
-
Run the following command in the app's directory to run your web app.
python run.py -
Go to http://0.0.0.0:3001/
Loads the messages and categories datasets -> Merges the two datasets -> Cleans the data -> Stores it in a SQLite database
Loads data from the SQLite database -> Splits the dataset into training and test sets -> Builds a text processing and machine learning pipeline -> Trains and tunes a model using GridSearchCV -> Outputs results on the test set -> Exports the final model as a pickle file
- template
- master.html: Main page of web app.
- go.html: Classification result page of web app.
- run.py: Flask file that runs app
- disaster_categories.csv: Data to process.
- disaster_messages.csv: A data set containing real messages that were sent during disaster events.
- process_data.py: A data pipeline to prepare message data from major natural disasters around the world.
- InsertDatabaseName.db: Database to save clean data to.
- train_classifier.py: A machine learning pipeline to categorize emergency messages based on the needs communicated by the sender so that you can send the messages to an appropriate disaster relief agency.
- classifier.pkl: Saved model
For fitting the data, I tried a RandomForestClassifier, DecisionTreeClassifier, and KNeighborsClassifier, and evaluated precision, recall, f1_score, and accuracy. I prioritized the model that performed best at the f1_macro score since the dataset is imbalanced (ie some labels like water have few examples). The best model was a DecisionTreeClassifier tunned with various hyperparameters and features.
precision recall f1-score support
related 0.85 0.86 0.86 5026
request 0.58 0.54 0.56 1184
offer 0.00 0.00 0.00 32
aid_related 0.67 0.64 0.66 2819
medical_help 0.38 0.31 0.34 533
medical_products 0.46 0.40 0.43 324
search_and_rescue 0.26 0.19 0.22 180
security 0.12 0.08 0.10 122
military 0.43 0.38 0.40 221
child_alone 0.00 0.00 0.00 0
water 0.69 0.60 0.64 465
food 0.76 0.73 0.74 770
shelter 0.63 0.61 0.62 566
clothing 0.54 0.51 0.52 99
money 0.37 0.29 0.33 170
missing_people 0.19 0.10 0.13 86
refugees 0.34 0.26 0.29 211
death 0.60 0.59 0.60 309
other_aid 0.32 0.25 0.28 890
infrastructure_related 0.19 0.16 0.18 413
transport 0.34 0.27 0.30 316
buildings 0.48 0.41 0.44 361
electricity 0.37 0.34 0.36 128
tools 0.03 0.02 0.03 43
hospitals 0.18 0.15 0.17 72
shops 0.00 0.00 0.00 23
aid_centers 0.12 0.08 0.10 75
other_infrastructure 0.18 0.13 0.15 290
weather_related 0.73 0.74 0.73 1762
floods 0.64 0.58 0.61 549
storm 0.65 0.70 0.67 578
fire 0.36 0.29 0.32 68
earthquake 0.79 0.79 0.79 577
cold 0.56 0.42 0.48 138
other_weather 0.22 0.19 0.20 343
direct_report 0.53 0.46 0.49 1308
micro avg 0.64 0.60 0.62 21051
macro avg 0.40 0.36 0.38 21051
weighted avg 0.62 0.60 0.61 21051
samples avg 0.52 0.50 0.46 21051
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent