From Thought To Cluster - Part I

A Machine Learning model is the outcome of a specific Machine Learning algorithm that has run on specific training data. The Machine Learning model is an entity that has already learned to produce a certain type of output according to the specific data it has “learned on”. This model is now ready to receive non-training data.

You can also imagine the Machine Learning algorithm as a building’s blueprints, and the data as blocks, cement and other building materials. The blueprints are just a plan of action and they don’t actually materialize until they meet the blocks or the cement.

Only after the building materials (training data) have been ordered according to the blueprints (run through the algorithm) will the building turn into a building (model).

The Machine Learning model is the heart of any Machine Learning System. You can think about it as the engine inside the car that is our giant information deriving machine.

These insights may be:
  - A prediction about the future values of the data
  - Some sort of description of the past
  - Suggestions about the future of the data
  - Or any other pattern that can be derived for the data

After reliable insights and information have been extracted and organized, they will usually be translated into insights. These insights are what we eventually want in order to improve decision making in the real world in organizations and businesses.

Basically, a Machine Learning System is the highest level view point of the input-output black box. It’s the part that holds inside the Machine Learning Model, input data cleaning, output handling system and all the other pieces.

The responsibilities of a Machine Learning System can be (all or some of) the following:

  - Accepting large amounts of input data
  - Preprocessing the input data
  - Transporting the processed data into the model
  - Running the Machine Learning Model
  - Post processing the information
  - Returning an output at the end

The general idea behind using a Machine Learning system is to automate the process of learning from data. Using Machine Learning Systems allows us to keep improving the learned insights and not have to reprogram the whole system every time we want to improve some more.

Though preprocessing of the input data is not always required, it is still in many times a big part of a ML system. This is always a point that is worth taking a few moments to think about.

Many times the format of the input data will be different from the format received by your ML system. At other times the input data may contain within it data bits that are unnecessary for the model. The process involving snipping, formatting and ordering the input data into the Machine Learning system is referred to as Data Cleansing.

(*The size of your input data may vary but in order for most ML systems to work, the input data will usually be similar to the training data.)

Just like the preprocessing of the input data to our Machine Learning system, the visualization of data is not an essential part of what makes a Machine Learning system a Machine Learning system.

With that being said, data visualization is still a crucial part of the whole information deriving process. It is oftentimes a big part of the responsibilities of a data scientist.

It basically means taking the outputs from the Machine Learning system and turning them into readable and understandable human insight. After all, what is an almighty Machine Learning system worth if we can’t understand anything that comes out of it???

Data visualization may also be (partially or fully) automated inside the boundaries of the Machine Learning System.

Data visualization can be outputting what we’ve learned into any one of the following:

  - Data tables
  - Graphs
  - Flow charts
  - Diagrams
  - Any other understandable way of relaying information