2. WHAT IS MARKET SENTIMENT ANALYSIS?
MARKET SENTIMENT IS THE OVERALL ATTITUDE OF
INVESTORS TOWARD A PARTICULAR SECURITY OR FINANCIAL
MARKET. MARKET SENTIMENT IS THE FEELING OR TONE OF A
MARKET, OR ITS CROWD PSYCHOLOGY, AS REVEALED
THROUGH THE ACTIVITY AND PRICE MOVEMENT OF THE
SECURITIES TRADED IN THAT MARKET.
MARKET SENTIMENT, ALSO CALLED "INVESTOR SENTIMENT,"
IS NOT ALWAYS BASED ON FUNDAMENTALS.
3. 1. MACHINE LEARNING APPROACH
Machine learning techniques that are applied in the field of sentiment
analysis can be divided as supervised(1.1) and unsupervised(1.2) learning
methods.
4. 1.1 SUPERVISED SENTIMENT
ANALYSIS.
Sentiment analysis is a method to understand public attitude toward a
topic/ product. (with machines algorithm to automatically classify/ learn
thousand of talk without having to read them manually).
This is how brokerage website works.
The idea is to learn and to understand what is “bullish” or “bearish
message.
Supervised learning is one that makes use of known dataset to make the
prediction of output result.
5. 1.1 TYPES OF SUPERVISED LEARNING-
Decision tree classifier. (1.1.1)
Classification tree. (1.1.1.1)
Regression tree. (1.1.1.2)
Rule based classifier. (1.1.2)
Probabilistic classifier.(1.1.3)
Naïve base classifier. (1.1.3.1)
Maximum entropy classifier. (1.1.3.2)
Linear approach. (1.1.4)
Support vector machine. (1.1.4.1)
Neural network. (1.1.4.2)
NOTE:- Classification tree. (1.1.1.1)+Regression tree. (1.1.1.2)= CART
6. 1.1.1. Decision tree classifier.
Decision trees can be visualized and are simple to understand and
interpret.
They require very little data preparation whereas other techniques often
require data normalization, the creation of dummy variables and removal
of blank values.
Decision trees can handle both categorical and numerical data whereas
other techniques are specialized for only one type of variable.
Decision trees can handle multi-output problems.
Decision trees can perform well even if assumptions are somewhat violated
by the dataset from which the data is taken.
7. 1.1.1.1. Classification tree.
Classification tree in which analysis is when the predicted outcome is the
class to which the data belongs. For example outcome of loan application
as safe or risky.
8. 1.1.1.2. Regression tree.
Regression tree in which analysis is when the predicted outcome can be
considered a real number. For example population of a state.
9. 1.1.2. Rule based classifier.
Rule-based classification is another type of supervised learning data
mining method.
This algorithm provides mechanisms that generate rules by concentrating
on a specific class at a time and maximizing the probability of the desired
classification.
10. 1.1.3. Probabilistic classifier.
In machine learning, a probabilistic classifier is a classifier that is able to
predict, given an observation of an input, a probability distribution over a
set of classes, rather than only outputting the most likely class that the
observation should belong to.
11. 1.1.3.1. Naïve base classifier
This classifier is based on Bayes theorem of probabilistic model.
In this we tried to estimate the probability of a text based on whether it
belongs to positive or negative class.
12. 1.1.3.2. Maximum entropy classifier
Maximum Entropy classifier is a probabilistic based classifier which belongs
to the exponential model class.
Principle of maximum entropy is used in this chapter and distribution
having largest entropy is chosen.
13. 1.1.4. LINEAR APPROACH
Linear classifier is one that partition a set of object into their respective
domain with a line, and partitioning with a curve is called as hyper plane.
1.1.4.1 Support vector machine
SVM is a supervised learning classifier widely used for classification and
regression analysis.
The basic idea of SVM is to determine linear separator in the search space
which can separate the different classes.
1.1.4.2 Neural network
This method is based on the neuron.
Neural network offer nonlinearity, input output mapping, adaptively and
fault tolerance.
14. 1.2 UNSUPERVISED SENTIMENT
ANALYSIS
Unsupervised learning has no explicit target output associated with input,
and it is learning through observation.
Famous approach in unsupervised learning is clustering.
15. 2. Lexicon based approach
For sentiment analysis lexicon based approach is robust that result in good
cross-domain performance.
This method is based on the assumption that the sum of the sentiment
orientation of each word makes contextual sentiment orientation.
16. 2.1. Dictionary Based approach
This approach use predefined dictionary of words where each word is
associated with a specific sentiment polarity strength.
Feeling of people such as happy, sad or depressed can be found out by
comparing word against lexicons from dictionaries.
17. 2.2. Corpus based approach
Corpus based approach try to find co-occurrence patterns of words to
determine their sentiments.
This approach is based on seeding list of opinion words and then find
another opinion words which have similar context.
This method is used to assign happiness factor of words depending on
frequency of their occurences in “happy” or “sad” blog post.
18. 3. Conclusion
There are various ups and downs in Indian stock market. In order to invest
money in stock market for purchasing the shares it is very essential for the
investors to predict the stock market condition. In India scenario Sensex
and Nifty are two major indicator for prediction of stock market condition.
For BSE (Bombay Stock Exchange) companies Sensex and for NSE (National
Stock Exchange) companies Nifty is used as an indicator of stock market
prediction. But the major problem for the investors are to predict the stock
market condition which depends upon regular checking and testing of
Sensex and Nifty prediction values. In order to allow this, in this paper we
have demonstrated sentiment analysis for stock market by fetching Sensex
and Nifty live server data values on different interval of time that can be
used for predicting the stock market status.