guidodc97 / dm

Repository for "Data Minig" project

Data Mining

Repository for Data Mining project.

Individual project consisting in building a classifier for Alzheimer's disease detection.

AY 2020/2021

Solution

Project has been completely developed in Python, using popular libraries for data analysis and machine learning such as pandas and scikit-learn.

The search for the best machine learning algorithm to adopt for the specific classification task consists of the following steps:

• Data preparation: which is articulated in:
  • Data cleaning: consists in the correction of incoherent data in the dataset where possible.
  • Data imputation: consists in replacing missing values with reasonably values.
• Model selection: the aim is to find out the most promising algorithm for the problem under study. This step is acomplished by performing a repeated stratified K-fold cross validation with inner grid search for finding best hyperparameters for algorithm.
• Hyperparameter tuning: once the model selection has been acomplished, the hyperparameter of the selected model can be tuned with a K-fold cross validation and an inner grid search for hyperparameters.
• Model evaluation: with the hyperparameter gfound at the previous step, it is possibile to train the model with them and to evaluate its performances. The evaluation phase is used again a K-fold cross validation.
• Prediction: finally, the model is retrained on the whole dataset with the specified hyperparameters foud during tuning phase and new instances can be predicted.

The best solution chosen is SCV with polinomial kernel, which gave an accuracy of 72,4% on test set.

Documentation & Code

Complete description of assignment can be found here.

Source code of project can be found here.