Data Science Assessment Task

📄 Take-Home Assignment: EUR/USD Weekly Direction Prediction

Objective

Your task is to build a machine learning model to predict the weekly price direction (up, down or no change) of the EUR/USD exchange rate.

The primary goal is not to maximize accuracy, but to demonstrate your understanding of ML principles and processes, including data handling, feature engineering, model design, and evaluation.

---

📊 Data Provided

You will be given a weekly time series dataset with the following daily variables:

• EUR/USD spot exchange rate
• US 2-Year Treasury Yield (USGG10YR)
• German 2-Year Bund Yield (GDBR10)
• EUR/USD implied volatility (EURUSDV1M)
• VIX (Volatility Index)

The dataset is provided in the file: Data_Scientist_Data.xlsx

---

📌 Your Tasks

1. Define the Prediction Problem

• Formulate the target variable as the next-week direction of EUR/USD (+1 if the exchange rate goes up, -1 if it goes down, 0 if it doesn't change)
• Explain your reasoning for framing the problem as classification/regression
• Your assignment should cover:
  • Exploratory Data Analysis
  • Feature Engineering
  • Model Development
  • Model Evaluation
  • Results and Interpretation
• Train at least two different ML models and compare their performance across key metrics
• Don't worry about accuracy score - you will not be assessed by the accuracy score at all. It's more about demonstrating good ML practices and reasoning
• Submit your Python Code or Jupyter Notebook

2. Software Development Extension: Interactive Dashboard

Objective

Develop a simple interactive dashboard to present and analyze the results of your EUR/USD weekly direction prediction model. This will allow a non-technical user to explore the data, run predictions, and understand the model outputs.

Requirements

Framework

• You may use Streamlit, Dash, or any similar Python web framework

Core Features

• Data upload: Allow a user to upload a CSV/Excel file in the same format as provided
• Model predictions: Run your trained model on the uploaded dataset (or on a subset of it)
• Visualizations:
  • Plot the historical EUR/USD spot rate and highlight weekly predicted vs actual directions
  • Show at least one model interpretability element (e.g., feature importance, SHAP summary, or coefficient weights)
• Performance summary: Display confusion matrix or classification metrics (precision/recall/F1) on test data

User Experience

• The dashboard should be easy to use and logically structured
• Include clear labels, titles, and short descriptions for plots/tables

Code Quality

• Structure your code so that the ML logic (data loading, feature engineering, training, prediction) is separated from the dashboard app code
• Include comments or docstrings where necessary

---

🔧 Deliverables

• Source code for the dashboard (Python files)
• Instructions for how to run the app locally (e.g., streamlit run app.py)
• (Optional) A requirements.txt file with dependencies

---

📋 Submission Instructions

Environment Setup

• Use pip for package management and dependency installation
• Create a requirements.txt file listing all necessary dependencies

Time Limit

• Complete this assignment within 2 days maximum
• Focus on demonstrating your approach and methodology rather than perfecting accuracy

Submission Process

• DO NOT commit your changes to GitHub
• Zip your entire project directory
• Send your completed assignment via email l

What to Include in Your Submission

1. All Python files (.py or .ipynb)
2. requirements.txt file
3. Brief documentation on how to run your code
4. Any saved model files

---