sal <- read.csv(file='C://Users//ckt//OneDrive - MUNI//Institutions//MU//ML Finance//2024//MidTerm//jobsal2024.csv')

# These are data of salaries for 'data scientist' jobs around the world
# The list of variables can be found in 'DataDescription.csv'
# The variable of interest is:
# salary        - annual salary in USD
# There are multiple groups of features:
# year          - for which the data is recorded
# Experience level
# Contract type
# Company size
# Employee residence
# Company residence
# Job title

# Your goal is to design a prediction model that will predict the salary in year 2024.
# Think about variables that you will use. You are strongly encouraged to
# create your own variables. 
# Your graded tasks are as follows:
# 1) Prepare a table with descriptive statistics and correlations for key variables (not necessarily all) that enter your model.
# 6 points

# 2) Use at least 4 models to predict salary. Using more complex models (random forest, xgb, elastic net) are more likely to result in maximum of 5 point here.
# 5 points

# 3) Evaluate models and suggest which model to use in future. Explain your recommendation.
# 5 points

# Remaining 14 points are rewarded based on the following:
# Apart from fulfilling the goal, you will be evaluated based on following criteria:
# - if you submit a working code (I load the data and run your code without errors. Think about reproducibility). (4b)
# - if you present Figures as well. (1b)
# - if you select different 'types' of models, but 4 is enough. (2b)
# - if you use multiple loss functions. (1b)
# - if you interpret your results correctly. (2b)
# - if you create your own variables and consider different transformations. (4b)

# - Upload this script with your working codes only.
# - Add text to it to describe your results
# - Use only prediction models we discussed during lectures and seminars.
# - You are allowed to use scripts 'functions' with functions that should be helpful.
# - You are allowed to use scripts - official ones - from classes.

# Tips
# - Do not use response/target variable salary as a feature - you would be surprised how often that happens.