Neural Networks for Health Technology Applications is a freely selectable course in Bachelor of Information Technology Degree Programme students at Helsinki Metropolia University of Applied Sciences. The following three case studies were completed during the course.
Case 1 - Heart Disease Classification
In this exercise we create an ANN to classify UCI Heart Disease -dataset.
Score: 14/15
Case 2 - Pneumonia
In this exercise we create CNN models to classify pneumonia in chest x-ray images.
Score: 19/20
Case 3 - Drug Reviews
In this exercise we build a model to predict numerical ratings from textual drug reviews written by patients. This is done using text processing and a 1D CNN.
Score: 15/15
Methods 1 - Data Manipulation
The aim of this exercise was to verify Anaconda3 software installation, create a Jupyter notebook and see that everything is running properly. Next, we try creating a pandas dataframe and get statistics from it, load data from a .csv file, iterate through rows and visualize a correlation matrix.
Methods 2 - Clustering Algorithms
In this exercise, we determine depot locations for a drone delivery system based on spatial data, using K-means clustering and agglomerative hierarchical clustering.
Methods 3 - Decision Trees
The aim of this exercise it to create a useful decision tree, visualize it and provide an accuracy estimate for it.
Methods 4 - Linear Regression
In this assignment, we learned how to conduct linear regression analysis with some feature engineering and awareness of the assumptions of linear regression.
Methods 5 - Logistic Regression
Using logistic regression to tackle a classification problem. Analyzing the predicrive quality of the model.
Methods 6 - Recommendation Engine
The aim of this exercise is to create a recommendation engine. The methods used and compared are K-Nearest Neighbors (KNN) and Singular Value Decomposition (SVD).
We learn how to deal with algorithms requiring excessive hardware resources.
Methods 7 - Text Analytics
In this exercise, we do a simple dictionary-based tone analysis and learn about natural language processing methods using nltk.
We visualize how the tone of the text develops between positive and negative.