-
web app for forest cover type prediction
Inspiration
The project is inspired by the paper Tackling Climate Change with Machine Learning, and I’m especially interested in automating afforestation. I want to work on a project related to afforestation, so I choose Forest Cover Type Dataset, which can be used to predict the forest cover type given the surroundings. This technique can be used to determine what species to plant in a particular area to increase the climate impact of afforestation.
What it does
The forest cover type predictor predicts what types of trees grow in a region based on the surrounding characteristics. The web app allows users to enter values for elevation, aspect, slope, horizontal and vertical distance to hydrology, hill shade at 9 am, 12 pm, and 3 pm, horizontal distance to fire points, wilderness area, and soil type to virtually plant a certain type of tree.
How we built it
The prediction model was built mainly with sklearn. The model used for this project is a RandomForest classifier. The web app was built with Flask, and part of the code was adapted from McGillAISociety Fall2020-Workshop3.
Challenges we ran into
It is a classification problem and the dataset is highly imbalanced. I need to balance the dataset to improve the performance of models. Since the dataset is very large, I need try different models and find an efficient one to predict the testing set.
Accomplishments that we're proud of
The testing accuracy is 91% and is very satisfying for me. This web app is the first web app I have built and it is a rewarding experience to see it works.
What we learned
I learned a lot about data preprocessing, for example, normalization, standardization, oversampling, and feature selection. I learned how to select a classifier to solve a certain problem and how to visualize a confusion matrix. I also learned basic tools for building a web app.
What's next for Forest Cover Type Prediction
I want to add more descriptive information about the wilderness areas and soil types so that users can learn more about the surroundings of the tree they plant.
