I applied the concepts from a machine learning course I took at Minerva University to work on real world datasets. Each project goes through a process of data cleaning, splitting into training and test set, modelling and evaluation. They also conclude with a short report on the veracity of the results and in some cases a comparison between the results from using different models.
Methods: K-Nearest Neighbours, Ridge Regression, Cross-Validation
Using data from the Loaning Club from 2011 to 2017 to predict which loans will be accepted and rejected based on an application and the largest loan amount to be successfully funded.
Methods: Linear Parameter Models, Ridge Regression, Cross-Validation
Using past temperature data at Yosemite village, predict the temperature at a specific time and day of the year, accounting for the daily, seasonal and yearly variations.
Methods: Linear Classifier, Principal Component Analysis, Cross-Validation, Logistic Regression, Linear Discriminant Analysis, Support Vector Machines, Kernels (Linear, RBF, Poly), Deep Learning (VGG16)
Using images of clothes, predict which images are woman's and man's clothing.
Methods: Hidden Markov Model, Expectation Maximisation
Given text, determine which language the text is written in and given an audio file, determine which speaker is speaking each portion of the audio.
Methods: Random forest, Cross-Validation
Given a patient's data before being admitted into the ICU, predict the mortality of the patient.