rajat1994/alzheimersml

Alzheimer's Machine Learning Project

AlzheimersML

This work aims to define the progression of Alzheimer's disease and explore the predictors for the disease by analyzing the functional magnetic resonance imaging (fMRI) images. The dataset is used for constructing a regression model mapping the features extracted from fMRI images to the Mini Mental State Examination (MMSE) score.

To run the script do the following:

• pip install the pip_dependencies.pip file to get the required dependencies you can also install the dependencies manually:
  • numpy
  • scikit-learn
  • scipy
• type the following command:

  python run.py fMRI_train.csv fMRI_test.csv
  

To perform analysis on the output, use the following command:

python analysis.py [name_of_output.csv] [name_of_test]

The information below was given to us for the assignment:

References [1] Gong, Pinghua, Jieping Ye, and Changshui Zhang. "Robust multi-task feature learning." Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 2012.

---

Training sample size : 800 Training data is in fMRI_train.xlsx. Each row is an example and each column is a feature. The features in each row are used to describe the corresponding example.

Test sample size : 200 Test data is in fMRI_test.xlsx.xlsx

---

Number of features : 285 These features are normalized so that each feature has a 0 mean and standard deviation of 1.

---

Labels : MMSE scores Labels of each patient (correspondingly their fMRI image) are the real-valued scores.

---

File format: For both training and test data files, the first row gives the feature names. The first column gives the example ID which is not a feature, and you should not use it when training your model. The second column gives the label for each image (or patient) as explained above, and the values are real numbers. This is the target variable that you want to predict based on other features of each patient.

Your task is to train a regression model using a machine learning method studied in our class. You should only build your model from the training data, and then test your model on the test data. You can either write your own codes to implement the method or download some existing machine learning packages. If you decide to use downloaded package, please state the source of your download so TA can understand. If you decide to use a downloaded package, you might need to adapt it to the given data. You are welcome to also try to figure out some other methods not studied in our class that perform better than our studied techniques. This will bring extra credits of 5 points additional to the 40 total points. You need to make sure a in-class-studied technique is used first to compare with your other methods.