#Need guidance: how to start a Machine Learning project without prior experience

1 messages · Page 1 of 1 (latest)

modest loom
#

I recently completed the Machine Learning specialization from DeepLearning.ai with Andrew Ng, but I haven’t yet had any hands-on experience. Now, I'm about to dive into my first machine learning project and could use some guidance.

The task I’ve been given involves data about companies and the partnerships they've made with one another. I need to build a model that can predict future partnerships for new companies.

Initially, I thought about using a recommendation system. However, since this isn’t about finding similar features but rather complementary ones, I’m not sure if this is the right approach. Would it make more sense to start with a different type of model or maybe a simpler deep learning experiment?

I’m fully committed to studying/practicing whatever is necessary to complete this project, but I’m currently feeling a bit lost on how to get started. Any guidance would be greatly appreciated!

echo steeple
#

This is a good starter into exploring different applications with very solid step by step guides for each model

#

Regarding your project, I think it's a bit of a fool's errand. Connections between companies are brokered via sales teams and business people. It is impossible to forecast with any certainty whether a company will agree to a partnership because of the variability that goes into forming those relationships.

modest loom
#

Thanks for the Keras documentation code examples, I will go through it. About the project, I did not explain more details because I did not want to make a long post but I should have specified that when I am talking about partnerships I am talking about agreements or collaborations that they may find depending on what service or product these companies may need or produce. Simple example, although it is not obvious, the company that makes machine learning software may be interesting for an old family company with a small production line. It is about matching possible synergies.

echo steeple
#

I still say that is a fool's errand. Your model would have to examine the existing market as well as anticipate business needs. The amount of logic and weighing is too complex for even an ensemble of models and wont guarantee results