# Challenge of Machine Learning 2020 — Data

来源：https://www.kaggle.com/competitions/challenge-ML-MIHE-UC3M/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

<h2>Provided Files</h2>
<ul>
<li><strong>data.csv</strong> - patient information from moth 0 to 24</li>
<li><strong>sampleSubmission.csv</strong> - a sample submission file in the correct format</li>
</ul>
<h2>Data description (data.csv)</h2>
This file contains all the available information of the 1739 patients during a two-year follow-up. Each row of this file is associated to a single patient and the columns have the following characteristics:


<ul>
<li><strong>Id</strong> - anonymous id. Every patient has its own unique Id</li>
<li><strong>test_sample</strong> - a boolean variable (value 0 or 1) indicating with a value of 1 that this patient will be included in the test data</li>
<li><strong>GENDER</strong> - Female or Male (0, 1) </li>
<li><strong>AGE</strong> -Patient age at month 0</li>
<li><strong>PTEDUCAT</strong> - Years of education</li>
<li><strong>APOE4</strong> - Number of this gene in the patient (0,1,2)</li>
<li><strong>FDG-PET derived measures: AngularLeft, AngularRight, CingulumPostBilateral, TemporalLeft, TemporalRight</strong> - FDG-PET measurements in different brain regions at month 0 (measure the presence of Fluorodeoxyglucose from PET data)</li>
<li><strong>Temporal diagnosis change: DXCHANGE</strong> - Categorical variable indicating a change in the patient diagnosis. Considering four states S0 (control patient), S1 (medium state), S2 (severe state) (1 = Stable:S0 to S0, 2=Stable:S1 to S1, 3=Stable:S2 to S2,  4=Conv:S0 to S1, 5=Conv:S1 to S2, 6 = Conv:S0 to S2, 7 = Rev:S1 to S0, 8 = Rev:S2 to S1, 9 = Rev:S2 to S0, -1 = Not available)</li>
<li><strong>Temporal evolution in neurophysiological and behavioural tests </strong> - For each time-stamp we have 5 entries: <strong>MMSE</strong>, <strong>RAVLT learning</strong>, <strong>RAVLT immediate</strong>, <strong>RAVLT perc forgetting</strong>,   <strong>FAQ</strong></li>
<li><strong>MRI summarized data </strong> - For each time-stamp we have average MRI data from 7 brain regions: <strong>Ventricles</strong>, <strong>Hippocampus</strong>, <strong>Whole Brain</strong>, <strong>Entorhinal</strong>,   <strong>Fusiform</strong>, <strong>MidTemp</strong> and <strong>ICV</strong></li>
<li><strong>Temporal evolution  of AV45-PET derived measures</strong> - For each time-stamp we have 7 AV45-PET measurements in different brain regions: <strong>Cerebellum Grey Matter</strong>, <strong>Whole Cerebellum</strong>, <strong>Eroded Subcortical </strong>, <strong>Frontal</strong>,   <strong>Cingulate</strong>, <strong>Parietal</strong> and  <strong>Temporal</strong></li>


</ul>

All features with temporal evolution are indicated with a flag <strong>t_X_</strong>, where X denotes the acquisition month (0, 3, 6, 12, 18, 24, 30, 36).

For each patient we have demographic information (age, gender, education level, genetic information, …) together with different medical tests performed periodically (month 0, 3, 6, 12, 18, 24) that provide FDG-PET and AV45-PET derived measures, neurophysiological and behavioural tests, MRI summarized data, and CSF (cerebrospinal fluid) biomarkers.

Besides, together this information, we have a scoring characterizing the Disease Progression (DP).

So, the final goal of this challenge is the estimation of the DP scoring at month 36 using all the patient information collected from months 0 to 24.

<h2>Prediction submission (sampleSubmission.csv)</h2>
This file is an example of the submission file format and, besides, it contains the results of the baseline approach. 
This file has to contain, for each patient, two columns: Id and Predictions. Note that only Id patients with flag 'test_sample'=1 will be considered for the testing evaluation</p>
<p>The submission file should contain a header and have the following format:</p>
<pre>Id,Predictions<br>1,1.0<br></pre>


Here, you can find a demo notebook to generate the baseline solution:

https://www.kaggle.com/vgverdejo/baseline