Discussing this on a high level, these are the following steps that usually happen in a machine learning process.
Gathering data: Data is first gathered, to be played around with and analyzed.
Data pre-processing: data is never clear to understand. This data is processed and converted to information before it's used.
Chose the right model/algorithm: There are various models or algorithms used for predicting the output. You need to decide which model is the best suited for your requirement.
Train the model: Once you've gathered the data, processed it, decided a model, you need to start training your model.
Test the model: once you've trained it, you have to test if it actually works or no. If it does well and good, if it does not, you have to re-train it. And hence this is an iterative process.
Tune the model: if your trained model cleared the testing stage, you need to start checking for accuracy. You need to check how accurate your model is. In the beggining, its never perfectly accurate and hence you need to tune for accuracy. This is done by adding many filters and tiny tiny inputs.
Deploy the predictions: Once all this is done, the final step is to deploy your predictions.