Hey, Do you really need Machine Learning?

Whenever you have some business problem which you are planning to solve with Machine Learning or there might possible you have existing solution and you are trying to solve/improvise the same with Machine learning.

First of all check,

Is this problem can be solved with non-ML approach?(with traditional programming)

If yes, then-

Is it necessary solve same problem with Machine Learning?

You can find out the answer to this question just by analyzing non-ML (or existing) and possible ML solution on below factors,

1. Quality:

You need to think,

how better the ML solution will be compared to non-ML solution (existing one) ?

If it’s just little bit better than existing solution,

Then it’s better to select the non-ML (or Existing) solution as best solution.

Developing ML solution approach would be costlier and might not be beneficial to business as it is adding just little bit value compared to non-ML solution.

Note: Some times small improvement may add more business value when we think about security or financial domain.

2. Development Cost:

How long project going to run?

Longer the project runs it will need more computational power and time of skilled resource which has direct impact on cost of project.

Is ML solution justifying increased cost?

If it’s large business, then in most of the time cost of ML solution is justifying.

But if business is small and then it’s difficult to justify the high cost of ML project.

According to Gartner, 85% ML projects fails.

3. Maintenance support:

What kind of maintenance required for this ML solution?

Is deployment support required?

Is production monitoring required?

Is model refining support required?

How long this maintenance will be required?

Again, getting answer to these questions will help you understand the estimated cost and time required for maintenance phase.

4. Resource availability:

Do you have ML skilled people?

If no, then,

Do you have HR people to recruit team?

Do you have trainers to train the resources on SOTA technologies?

Understanding business problem is very important responsibility of data scientist and project management team. It will help you to get go ahead for machine learning project.