Practice: ML Algorithm Selection

# Practice: classify AI problems by their ML approach

problems = [
    {
        "name": "Credit card fraud detection",
        "has_labels": True,
        "output_type": "category",
        "data_type": "tabular",
    },
    {
        "name": "Group website visitors into segments",
        "has_labels": False,
        "output_type": "groups",
        "data_type": "tabular",
    },
    {
        "name": "Classify photos as indoor or outdoor",
        "has_labels": True,
        "output_type": "category",
        "data_type": "images",
    },
    {
        "name": "Train a robot to pick up objects",
        "has_labels": False,
        "output_type": "actions",
        "data_type": "sensor_data",
    },
    {
        "name": "Predict tomorrow's temperature",
        "has_labels": True,
        "output_type": "number",
        "data_type": "tabular",
    },
]

def recommend_approach(problem):
    if not problem["has_labels"] and problem["output_type"] == "actions":
        return "Reinforcement Learning"
    elif not problem["has_labels"]:
        return "Unsupervised Learning"
    elif problem["data_type"] == "images":
        return "Supervised Deep Learning (CNN)"
    elif problem["output_type"] == "category":
        return "Supervised Classification (ML)"
    elif problem["output_type"] == "number":
        return "Supervised Regression (ML)"
    return "Needs more information"

print("Problem -> Recommended ML Approach")
print("=" * 60)
for problem in problems:
    approach = recommend_approach(problem)
    print(f"  {problem['name']}")
    print(f"  -> {approach}")
    print()