wyshanw Goto Github PK

bm_final_project

In this project, we aim to build regression models based on a set of demographic variables to estimate county-level crime rates. After an exploratory analysis of the variables on their distributions and correlations, we derived more meaningful variables by manipulating the existing ones, removed outlier values, and implemented several variable selection methods. Using the selected variables, we trained a linear regression model, elaborated on the model by adding several interactive terms, and did cross-validations. All 3 resulting models have achieved a good estimation of the training set. We selected the third model (adjusted R-square: 0.542, RMSE: 16.46, RMSPE: 11.90) as the final model, as it has the best prediction on testing set. Further studies can be done on correcting the dataset using external data sources, as well as using more sophisticated non-linear models.

data-homicides

The Washington Post collected data on more than 52,000 criminal homicides over the past decade in 50 of the largest American cities.

data_wrangling_i

data_wrangling_ii

example_repo

final_project

iteration_and_list_columns

linear_models

p8105_hw1_yw3772

p8105_hw2_yw3772

p8105_hw3_yw3772

p8105_hw4_yw3772

p8105_hw5_yw3772

p8105_hw6_yw3772

simulation

viz_and_eda

writing_functions

wyshanw.github.io