Machine Learning

What is Data Analysis

Let see Data Analytics pipeline

Step 1: Data Acquisitions

Step 2: Data Preparation

Step 3: Data Representation

Step 4: Data Representation

Step 5: Data Analytics

Step 6: Results Interpretations

in this post we are going to Data analysis on twitter data

What kind of Analytics can be done in Twitter?

• Counting. Real-time counting analytics such as how many requests per day, how many sign-ups, how many times a certain word appears, etc.

• Correlation. Near-real-time analytics such as desktop vs. mobile users, which devices fail at the same time, etc.

• Monitoring. Monitoring the customer opinions about a brand.

• Research. More in-depth analytics that run in batch mode on the historical data such as what features get re-tweeted, detecting sentiments, etc.

• Network Analysis: ego network analysis, monitoring followers growth, community analysis

• Sentiment Analysis: tracking events and hot topics, trending. monitoring customer opinions about the product

• Other: tweet engagement (top popular tweets)

What is Spark?

Apache Spark is a framework for developing distributed computing applications.

As research project at the University of California, Berkeley.

Features of Apache Spark

• Speed (in-memory computations)

• Supports multiply languages (Java, Scala, Python, R)

• Advanced Analytics (SQL queries, Streaming data, Machine learning and Graph algorithms )

System requirements

• Java 8+

• Scala 2.11.x

• Pyhotn 2.7+

• 8GM RAM and 8-16 cores CPU

• read more

Who uses Spark and Why?

Data Scientist:

• Analyze and model the data to obtain insights of the data;

• Transforming the data into a useable format

• Statistics, machine learning, SQL

• Advanced analytics

Engineers:

• Develop a data processing system or applications

• Monitor, inspect and tune the applications

What is Spark SQL

Allows:

• load the data in .csv, .json and .parquet file format

• relational queries expressed in SQL, Scala

• use of SchemaRDD (abstract table)

DataFrame, Dataset

A Dataset is a strongly typed collection of domain-specific objects.

Each Dataset has an untyped view called a DataFrame (a Dataset of

Row)

Two types of operations on Datasets:

• transformations (e.g., map(),filter(),select(),aggregate(), etc.)

• actions (e.g., count(),show(), etc. )

Getting the data/Load the data to Spark (SQLContext)

Understanding the data:

• calculating basic statistics

• making histograms

Cleaning the data:

• filtering data

• dealing with missing, incomplete data

Feature extraction:

• Dealing with categorical data

Clustering (using Mllib)

Saving the data to:

• files

• MongoDB database

Visualizing Data (Zeeplin or d3.js)