Step 1: Collect Data
Given the problem you want to solve, you will have to investigate and obtain data that you will use to feed your machine. The quality and quantity of information you get are very important since it will directly impact how well or badly your model will work. You may have the information in an existing database or you must create it from scratch. If it is a small project you can create a spreadsheet that will later be easily exported as a CSV file. It is also common to use the web scraping technique to automatically collect information from various sources such as APIs.
Sign up for your weekly dose of what's up in emerging technology.
Step 2: Prepare the data
This is a good time to visualize your data and check if there are correlations between the different characteristics that we obtained. It will be necessary to make a selection of characteristics since the ones you choose will directly impact the execution times and the results. You can also reduce dimensions by applying PCA if necessary.
Additionally, you must balance the amount of data we have for each result -class- so that it is significant as the learning may be biased towards a type of response and when your model tries to generalize knowledge it will fail.
You must also separate the data into two groups: one for training and the other for model evaluation which can be divided approximately in a ratio of 80/20 but it can vary depending on the case and the volume of data we have.
At this stage, you can also pre-process your data by normalizing, eliminating duplicates, and making error corrections.
Step 3: Choose the model
There are several models that you can choose according to the objective that you might have: you will use algorithms of classification, prediction, linear regression, clustering, i.e. k-means or K-Nearest Neighbor, Deep Learning, i.e Neural Networks, Bayesian, etc.
There are various models to be used depending on the data you are going to process such as images, sound, text, and numerical values. In the following table, we will see some models and their applications that you can apply in your projects:
|Logistic Regression||Price prediction|
|Fully connected networks||Classification|
|Convolutional Neural Networks||Image processing|
|Recurrent Neural Networks||Voice recognition|
|Random Forest||Fraud Detection|
|Reinforcement Learning||Learning by trial and error|
|Generative Models||Image creation|
|k-Nearest Neighbors||Recommendation systems|
|Bayesian Classifiers||Spam and noise filtering|
Step 4 Train your machine model
You will need to train the datasets to run smoothly and see an incremental improvement in the prediction rate. Remember to initialize the weights of your model randomly -the weights are the values that multiply or affect the relationships between the inputs and outputs- which will be automatically adjusted by the selected algorithm the more you train them.
Step 5: Evaluation
You will have to check the machine created against your evaluation data set that contains inputs that the model does not know and verify the precision of your already trained model. If the accuracy is less than or equal to 50%, that model will not be useful since it would be like tossing a coin to make decisions. If you reach 90% or more, you can have good confidence in the results that the model gives you.
Step 6: Parameter Tuning
If during the evaluation you did not obtain good predictions and your precision is not the minimum desired, it is possible that you have overfitting -or underfitting problems and you must return to the training step before making a new configuration of parameters in your model. You can increase the number of times you iterate your training data- termed epochs. Another important parameter is the one known as the “learning rate”, which is usually a value that multiplies the gradient to gradually bring it closer to the global -or local- minimum to minimize the cost of the function.
Increasing your values by 0.1 units from 0.001 is not the same as this can significantly affect the model execution time. You can also indicate the maximum error allowed for your model. You can go from taking a few minutes to hours, and even days, to train your machine. These parameters are often called Hyperparameters. This “tuning” is still more of an art than a science and will improve as you experiment. There are usually many parameters to adjust and when combined they can trigger all your options. Each algorithm has its own parameters to adjust. To name a few more, in Artificial Neural Networks (ANNs) you must define in its architecture the number of hidden layers it will have and gradually test with more or less and with how many neurons each layer. This will be a work of great effort and patience to give good results.
Step 7: Prediction or Inference
You are now ready to use your Machine Learning model inferring results in real-life scenarios.