Machine Learning

7 Steps Of Machine Learning

1. Data Gathering
2. Data Preparation
3. Model Selection (Apply appropriate simple algorithms or math)
4. Training (Applying Intel)
5. Evaluation (Test and Tweaking)
6. Hyper Parameter Tuning (Tune inputs to increase the accuracy of the output)
7. Prediction (Final o/p which gives useful value)

Types of Machine Learning

Supervised:

Some algos which follows supervised learning such as => Regression, Decision Tree, Random Forest, KNN, Logistic Regression algorithms.

How these algo works: target required output (dependent variable) that has to be predicted using some set of input independent variables. Putting together, identify a math equation to get the desired output. Many times, this is not one time process, apply more iterations to identify the best equation with high accuracy.

Example: 

let say, we get the data for a X person where he/she is at (long, lat), some Y time.

Here, we have 3 inputs => long, lat, Y time for a X person, where X is unique which provides data of those 3 variables.

Now, if we have to find out where a person spends most of the time or some top 3 rankings on the time spending.

So, to identify that, we need some data source map for the long and lat. For this, google map is the best. Here I am just trying to explain how we can apply supervised learning, will come up with actual detail steps to identify the output......more info coming soon....

Unsupervised:

In this, there is no specific output instead data clustering or grouping is done and in later stages these clusters helps in locating the input variables to appropriate clusters.

Popular algos: Apriori algorithm, K-means

Reinforcement:

In this, application or any software system is trained to use the past history data to apply or reinforce required math to get the desired output. Simple example in any app is the basic validation which is applied to generate the right output for the business.

In detail, we can write complex algo to design how an app should behave against the data change keeping the history knowledge.

Popular algo: Markov Decision Process

This is commonly used without we knowing that ML is used, in any app decisions are made using the past data to generate better and desired output.

Popular technologies

R

Python - my focus

Scala 

Developed by Martin Odersky, one of the powerful language targeted mainly focusing the scale-ability and to run on JVM. This is not Java but OOP and functional programming language.
To work with Big Data - Hadoop Spark framework, Scala is needed.

Scala originated from Java to make it more simpler and have less code for the developers. Java by itself is a wonderful language but since the world is changing, change was needed and Scala raised.

Scala is the language which will change the way you think if you really get into it.

Julia

More to come or they exists (explore)