-1
$\begingroup$

I learned that 87% of machine learning projects fail due to these five pitfalls:

  1. the scope of the project is too big;
  2. the project’s scope increased in size as the project progressed—e.g., scope creep;
  3. the model couldn’t be explained, hence there was lack of trust in the solution;
  4. the model was too complex; and
  5. the project solved the wrong problem.

And I learned that rules and heuristic approaches may be a better choice than machine learning, since the development of machine learning takes more time(and more costly) and its explanation is also harder.

Then I wonder when we should just use rules and heuristics and when we should bravely take the machine learning approach?

References:

  1. Why Ai/Data Science Projects Fail
$\endgroup$

1 Answer 1

0
$\begingroup$

Machine learning is an approach to (1) learn (2) complex patterns from (3) existing data and use these patterns to make (4) predictions on (5) unseen data.

Then there are 9 use cases here:

  1. Learn: the system has the capacity to learn

  2. Complex patterns: there are patterns to learn, and they are complex ML

  3. Existing data: data is available, or it’s possible to collect data

  4. Predictions: it’s a predictive problem

  5. Unseen data: Unseen data shares patterns with the training data

  6. It’s repetitive

  7. The cost of wrong predictions is cheap

  8. It’s at scale

  9. The patterns are constantly changing

However, even if ML can’t solve your problem, it might be possible to break your problem into smaller components, and use ML to solve some of them.

References:

  1. Designing Machine Learning Systems
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .