| **You can download the initial exercise file from [link](https://quera.org/contest/assignments/103149/download_problem_initial_project/356094/).** |
| ------------------------------------------------------------------------------ |
Welcome to the underground world of "Albuquerque"! Walter White has recently realized that having the purest chemical formula is not enough to conquer the market; he has figured out that the real power lies in accurately predicting customer behavior and psychology!
But a massive crisis has struck the empire! The Drug Enforcement Administration (DEA), led by Hank Schrader, is using artificial intelligence algorithms to identify consumption patterns across the city. Meanwhile, Heisenberg discovered that his lawyer, Saul Goodman, was guessing consumer behavior entirely by chance and merely by looking at the appearance and clothes of people in his office waiting room! Heisenberg fired Saul Goodman from the data analysis department out of pure anger and now senses the danger of his empire collapsing.
Now, Heisenberg wants one thing from you: build a system that uses valid psychological tests (NEO-FFI) and demographic information to predict the consumption class of 5 specific products for each individual. He has also collected a dataset of various people and provided it to you.
The extracted information has been provided to you in the form of two files (train and test). Each row of this dataset represents a person's profile. The details of the columns in this dataset are as follows:
| **Column Name** | **Description** |
| --------------------------- | --------------------------------------------------- |
| **ID** | Unique identifier for each person |
| **Age** | Age group of the individual (normalized) |
| **Sex** | Gender of the individual |
| **EducationLevel** | Education level of the individual |
| **Country** | Country of residence |
| **Background** | Ethnicity of the individual |
| **EmotionalStabilityScore** | Neuroticism score from the personality test |
| **SocialEnergyScore** | Extraversion score from the personality test |
| **OpennessScore** | Openness to new experiences score |
| **CooperationScore** | Agreeableness score |
| **SelfDisciplineScore** | Conscientiousness score |
| **ImpulseControlScore** | Degree of impulsive and thoughtless decision-making |
| **NoveltySeekingScore** | Sensation seeking and risk-taking index |
In the training file, in addition to the above columns, there are also 5 Target columns that you must predict for the test data. The values of these 5 columns include the numbers `0`, `1`, and `2`, which indicate the time frame of consumption:
+ **Class 0 (No/Distant Use):** Never consumed.
+ **Class 1 (Near Use):** Consumed in the past year.
+ **Class 2 (Regular Use):** Regular consumption.
These 5 targets are:
`Marijuana`, `LSD`, `Mushroom`, `Psychotropic`, `Ex`
### **Output**
To evaluate your program, you must predict the consumption class (numbers 0, 1, or 2) for the 5 target columns in the `test.csv` data. Your output must be a CSV file containing the `ID` column and the 5 predicted columns.
Code snippet
```
ID,Marijuana,LSD,Mushroom,Psychotropic,Ex
1,0,0,1,0,2
2,2,0,0,1,0
3,1,1,0,0,0
4,0,0,0,2,1
...
```
### **Evaluation Method**
Your model's performance in this problem is measured using the **Macro F1** metric. Since we are dealing with a multi-target classification problem, the judging system first calculates the _Macro F1_ score for each of the 5 targets separately. For your model to earn points in any target, it must be able to surpass the baseline performance threshold (66% accuracy). Scores above this threshold are scaled and form your final score in the range of 0 to 100.
The formula for calculating the final score is as follows:
$$Score = \sum_{i=1}^{5} 20 \times \min\left(1, \max\left(0, \frac{\text{MacroF1}_i - 0.50}{0.50}\right)\right)$$
> **If the model's performance in a target is less than 0.66, it will not receive any points for that section.**