Portfolio

For my graduate machine learning class, I created a neural network that makes predictions based off of the Austin Police Department's Racial Profiling dataset for 2019
https://github.com/24juangonzalez/Neural-Network-Racial-Profiling
0 forks.
0 stars.
0 open issues.

I coded the AdaBoost Ensembles using Python. AdaBoost, or Adaptive Boosting, was used for my ensembles as it adapts to the errors of the previous weak hypotheses. I imported the necessary libraries using the sklearn library. I used pandas to read the csv and replaced my selected Categories as numbers: White 1, Black 2, Latinx 3, Asian 4, American Indian 5, Middle Eastern 6, Hawaiian 7, unknown 8, and nans as 0 for the race description. I assigned other categories numbers since my code in Python was not reading strings properly. I made X as race description and Y as race known. Next, I created X and Y into arrays and looked up how to use iloc. For X, I selected all the rows and only the first column minus the other 5 columns as shown by 0:-5 in my X variable. Repeated this for the Y variable, selecting the second and the last 6th column. The Y variable is race known as it changes the whole argument if the cop knew the individual’s race before pulling them over. Using this method, I achieved an accuracy rate of 91.7%.
https://github.com/24juangonzalez/ADABoost_Ensemble
0 forks.
0 stars.
0 open issues.

For my decision trees, I did this in R studio. I first had to import the libraries I needed. I then imported my csv file using read.csv in R. I made sure to import the categories that I needed. These categories are APD_RACE_DESC,RACE_KNOWN, Person Search YN, Search Based On, Search Found , and Reason for Stop. I had to split my data into test and training data, so I used a 70-30 split ratio. My formula for this was that the APD_RESC_DESC was equal to the sum of all the other categories added together. I then created the model using the training data and APD_RACE_DESC as my variable. For the prediction model, I used the formula I created and the test training data to determine the accuracy which was at a low 43%. The accuracy of the model was not as high as I would’ve liked it probably due to the way the model output was designed. My decision trees my model created came out nicely and labeled all the categories along with various nodes and the sample size.
https://github.com/24juangonzalez/Decision_TreesML
0 forks.
0 stars.
0 open issues.

Short Python code for a land calculator. Converts square feet to acres in decimal and whole form
https://github.com/24juangonzalez/Land-Calculator
0 forks.
0 stars.
0 open issues.

Code creates a calculator based on BMI, YMCA, and Navy weight classifications for Male and Female.
https://github.com/24juangonzalez/BMI-YMCA-NAVY-Calculator
0 forks.
0 stars.
0 open issues.