1. email (called ham) that reason we choose

1.
For Spam Detection

Q.1 How will you choose the right algorithm? (With
justification).

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

Ans: First we need to consider a goal of the system.
What are you trying to get out of this? Do you probability the email is a spam
or non spam (called Ham). If you trying to predict spam email, then you look
into supervised learning algorithm. If not predict the target value then look
into unsupervised learning algorithm. If you choose the supervised learning
what’s targeted value. If targeted value like discrete value then it’s comes
into the classification and if targeted value in a numeric form then it’s comes
into the regression. If you are not trying to predict the targeted value than
it’s comes into unsupervised learning algorithm.

Justification: In spam detection system, it’s
classified into two classes like spam email or non spam email (called ham) that
reason we choose supervised learning algorithm. Classification comes into the
supervised learning algorithm. In supervised learning we choose Naive Bayes
classifier algorithm to predict spam email or non spam email.

Q.2 Write in detail about the different steps that
you will perform to develop the machine learning application.

Ans: There are following steps to develop machine
learning application:

1.      Collect
data: Collect the sample by scraping a website and extracting data or you could
get information from different website. To save some time and effort, you could
use publically available data.

2.      Training:
Firstly parse each email into its constituent’s tokens than generate a
probability for each token. After then store spamminess value to the database.

3.      Filtering:
For each message, firstly scan the all message for the next tokens .query the
database for spamminess after than calculate accumulated message probabilities.

4.      Test:
After filtering the message test email are spam emails or non spam email
(called -Ham).

5.      Use
it: After developing application, if the entire previous step worked as
expected developer then encounter new data.

Training and testing data: Train the algorithm what
a spam email looks like or non spam email look like. Likely you have all the
previous emails have been marked as spam by customers. Also test accuracy of
the spam filter. One idea would be test on same data that can be used for
training. However lead the major problem called over fitting.

Justification: we use 70 percentage of the for
training and 30 percentage data for testing than we avoid over fitting problem.
It is important to mix spam email and non spam email data into data sets.

2.
Stock Market Prediction

Q.1 how will you choose the right algorithm? (With
justification).

Ans: First we need to consider a goal of the system.
What are you trying to get out of this? Do you predict the future stock market
if you trying to predict stock price, then you look into supervised learning
algorithm. If not predict the target value then look into unsupervised learning
algorithm. If you choose the supervised learning what’s targeted value. If
targeted value like discrete value then it’s comes into the classification and
if targeted value in a numeric form then it’s comes into the regression. If you
are not trying to predict the targeted value than it’s comes into unsupervised
learning algorithm. Cluster technique use in unsupervised learning algorithm.

Justification: In stock market prediction system, we
use support vector machine because one of the most suitable algorithm for time
series prediction. In stock market prediction output comes into the numeric
form that mean it’s comes into regression and predict targeted value. That
reason we choose supervised learning algorithm.

Q.2 Write in detail about the different steps that
you will perform to develop the machine learning application.

Ans: There are following steps to develop machine
learning application:

1.      Collect
data: For data collection bank finance has been used to fetch three years
historical data for several stocks.

2.      Prepare
the input data: In Prepare input data redundancy data can be reduced from the
database.

3.      Training:
training set is foremost as it used to compute the gradient and updating the
training data set.

4.      Test:
After training the data test stock market data it is error free or not in stock
market prediction application.

5.      Use
it: After developing application, if the entire previous step worked as
expected developer then encounter new data.

Training and testing data: We will take bank like
state and central bank close price for last 2 months. We are taking nearest
close price for data consistency. Accurate data is very important, as even
number of data can cause regression algorithm function to change significantly.

Out of this data we will treat the first 40 days
data for training data and last 20 days data for the test data where we will
check how close prediction made by regression algorithm to the actual value.
Coefficient of determination is 0.85 that mean state bank and central bank
nifty is 85% correlated.

3.
Recommendations

Q.1 How will you choose the right algorithm? (With
justification).

Ans: First we need to consider a goal of the system.
What are you trying to get out of this? Do you predict the future rating of
system if you trying to predict rating, then you look into supervised learning
algorithm? If not predict the target value then look into unsupervised learning
algorithm. If you choose the supervised learning what’s targeted value. If
targeted value like discrete value then it’s comes into the classification and
if targeted value in a numeric form then it’s comes into the regression. If you
are not trying to predict the targeted value than it’s comes into unsupervised
learning algorithm. Cluster technique use in unsupervised learning algorithm

Justification: In Recommendation system, it’s
classified into two classes like recommend or non non recommended that reason
we choose supervised learning algorithm. Classification comes into the
supervised learning algorithm. Recommendation system mostly used in Baysian and
decision tree algorithm.

 

Q.2 Write in detail about the different steps that
you will perform to develop the machine learning application.

Ans: There are following steps to develop machine
learning application:

1.      Collect
data: For data collection from different website has been used to fetch three
month historical data for several rating.

2.      Prepare
the input data: In Prepare input data redundancy data can be reduced from the
database.

3.      Training:
training set is foremost as it used to compute the gradient and updating the
training data set.

4.      Test:
After training the data test recommendation system data it is error free or not
.

5.      Use
it: After developing application, if the entire previous step worked as expected
developer then encounter new data.

Training and testing data: We will take movie like
jay ho and titanic movie close rating for last 2 months. We are taking nearest
close rating for data consistency. Accurate data is very important, as even
number of data can cause classification algorithm function to change
significantly.

Out of this data we will treat the first 80 days
data for training data and last 20 days data for the test data where we will
check how close prediction made by classification algorithm to the actual
value.

 

.
4. Automated Breaking

Q.1 how will you choose the right algorithm? (With
justification).

Ans: First we need to consider a goal of the system.
What are you trying to get out of this? Do you predict the situation? If you
trying to predict automated break, then you look into supervised learning
algorithm. If not predict the target value then look into unsupervised learning
algorithm. If you choose the supervised learning what’s targeted value. If
targeted value like discrete value then it’s comes into the classification and
if targeted value in a numeric form then it’s comes into the regression. If you
are not trying to predict the targeted value than it’s comes into unsupervised
learning algorithm. This three terms not use then we use reinforcement
algorithm.

Justification: In Automated breaking system, it’s
not uses the classification regression and cluster because it is not comes into
supervised learning and unsupervised learning algorithm. It is comes into the
reinforcement algorithm. In reinforcement targeted value is given but do not
say output is wrong or Wright.

Q.2 Write in detail about the different steps that
you will perform to develop the machine learning application.

Ans: There are following steps to develop machine
learning application:

1.      Collect
data: Collect the data from previous history.

2.      Prepare
the input data: In Prepare input data redundancy data can be reduced from the
database.

3.      Training:
training set is foremost as it used to compute the gradient and updating the
training data set.

4.      Test:
After training the data test Automated breaking system data it is error free or
not .

5.      Use
it: After developing application, if the entire previous step worked as
expected developer then encounter new data.

Training and testing data: Out of this data we will
treat the first 80 days data for training data and last 20 days data for the
test data where we will check how close prediction made by reinforcement
algorithm to the actual value. Reinforcement algorithm do say output.

 

 

x

Hi!
I'm Barry!

Would you like to get a custom essay? How about receiving a customized one?

Check it out