Buzz Prediction on Twitter

Description:

There are two different datasets for Regression and Classification tasks. Right-most column in both the datasets is a dependent variable i.e. buzz.

Data description files are also provided for both the datasets.

Deciding which dataset is for which task is part of the project.

Read data into Jupyter notebook, use pandas to import data into a data frame.

Preprocess data: Explore data, check for missing data and apply data scaling. Justify the type of scaling used.

Regression Task:

Apply all the regression models you've learned so far. If your model has a scaling parameter(s) use Grid Search to find the best scaling parameter. Use plots and graphs to help you get a better glimpse of the results.

Then use cross-validation to find average training and testing score.

Your submission should have at least the following regression models: KNN regressor, linear regression, Ridge, Lasso, polynomial regression, SVM both simple and with kernels.

Finally, find the best regressor for this dataset and train your model on the entire dataset using the best parameters and predict buzz for the test_set.

Classification Task:

Decide about a good evaluation strategy and justify your choice.

Find best parameters for the following classification models: KNN classification, Logistic Regression, Linear Support Vector Machine, Kernelized Support Vector Machine, Decision Tree.

Which model gives the best results?

Deliverables:

Submit IPython notebook. Use markdown to provide inline comments for this project.

Rename notebook with your group number and submit only one notebook. Before submitting, make sure everything runs as expected. To check that, restart the kernel (in the menubar, select Kernel > Restart) and then run all cells (in the menubar, select Cell > Run All).

Visualization encouraged.