How to manage Machine Learning/Deep Learning project?

Source: Deep Learning on Medium


Recently, I’m getting many questions about the methodology that we used. We use CRISP-DM as everyone else. So what is this CRISP-DM. I’ll try to explain briefly. CRISP-DM is Cross Industry Standart Process for Data Mining.

CRISP-DM Approach

In seperately;

1. Business Understanding

2. Data Understanding

3. Data Preparation

4. Modeling

5. Evaluation

6. Deployment

This cycle continues with this sort. Firstly, we apply Business Understanding. What exactly is your business needs, what we want from us? We look for answers to these questions, Business Understanding and Data Understanding should be evaluated together. The first two parts are the subject of Data Science. Our role begins with Data Preparation. I want to tell you with data preparation, Where and how to get data? When we decide how to get data, we need to crop/clean the data in a way that we can use. It will take us a long time to adapt the data to our ML/DL model. In this part, we must be patient :)

The ML/DL subjects we read always correspond to the modelling. The main issues are prediction, classification, clustering, ARM, Reinforcement Learning, Natural Language Processing, Deep Learning and etc. We use all these algorithms in modelling. In later times, I will discuss all these algorithms in more detail.

We want to find hot to evaluate this model in the Evaluation section. Also we compare the models that we used in this section. The decision of which model we will continue with is based on the accuracy rate or any other criteria that you determine.

At the end of the project, you deploy your project to the server.