Process Map for Implementing Machine Learning Projects
In accordance with what I learned in the MIT program, I have created this diagram that visually outlines the steps to carry out an ML project. I hope it proves helpful to everyone and that we can use it as an additional tool to apply what we've learned. I would also appreciate any feedback on whether there are any concepts missing from this Process Map, which could be an area for improvement, and I would gladly send an update.
By utilizing Machine Learning, we make decisions based on predictions of real-world event behavior patterns. In my opinion, the first phase of ML, data understanding, is crucial in the field we are interested in, and I believe we can also explore the problem from the perspective of the processes within the entity where it occurs, whether it's a company, domain, or another field of study.
Creating a detailed map of each element and connection within the problem of interest and subjecting the data to algorithms to discover common patterns, dependent variables, independent variables, and predictive models is how we apply the second phase of ML: prediction.
After defining the model for predicting our target variable, we obtain data derived from the forecasts. We use this data for the third phase of ML, which is decision-making. To do this, we take into account the type of scenario, considering the dynamism of the environment and the amount of data available. Once we make decisions to implement in our business or field of study, we collect the resulting data for further analysis.
Finally, in the fourth phase of ML, causal inference, we evaluate whether these results were satisfactory and what factors had the greatest impact on the performance of our model. As we improve the predictive model, we enhance the entire chain of processes in the machine learning project we are implementing. This final phase allows us to understand the causes of the success or failure of policies implemented based on the data, algorithms, models, and methods learned in this program.
The concept that stands out the most to me is that we will never have a perfect model or be completely certain that decision-making will be 100% accurate. However, the more tools we are familiar with and the better we become as "data artists," the closer we get to success in implementing machine learning projects.
领英推荐
References: