matplotlib.txt

import matplotlib.pyplot as plt
plt.scatter(year,pop)
plt.show()

Histogram - distubtion of varaible. Divide into bins. 
            Hieght of the bar is the number of each bins

import matplotlib.pyplot as plt
plt.hist(x,bins=10)
plt.show()

#Label the axes
plt.xlabel("Year")
plt.ylabel("Population")
plt.title("Population growth")
plt.ytick([0,2,4],[0,4B])
#Log plot 
plt.xscale('log')
#Add Text
plt.text(1550, 71, 'India')
#Add Grid
plt.grid(True)

Pandas
======
1. select column: brics["country"] // Series
2. select column: brics[["country"]]
3. select row brics[1:4]
4. loc - select data based on labels. brics.loc[["RU","India"]]
5. loc([:],[[RU,India]]
5. iloc - brics.iloc[[1,2,3]]
6. brics.iloc[[1,2,3][:]]

#Comparison operator

abline - mean/median (pandas)
tapply (min,df)




#
Build models
Aquire, Refine, Transform, Explore,Model,
Model selection, 
Communicate the solution.


Data Space
Feature Space
Model alg try to apply (model algorithm)
Model alg + Parameter -
Search the hypotisis space.
Classes (c1,c2)
Features
Observations:1000
Do PCA. Pc1,Pc2

1. Train a model
2. Predict the porb of each Classes
3. Sort the model via 5 fold cross validationn
4. Show the decision boundires
5. Add SVM Linear, Decision Tree, KNN. 
6. Search Model
7. https://www.safaribooksonline.com/library/view/strata-hadoop/9781491958551/video293166.html
8. 


David Chapper
1. 10 Records - Columns
2. Data - 
   Machine learning alg - look for pattern data
   Output : Model - Able to recognize pattern
   Model - predict. Does this fit your data. 
3. Lots of data.
4. Compute power. 
5. Machine learning process
6. How can I predict if a new record is frudelent. 
7. Is the data predicitive. 
8. How would I know when I am done.

process
=======
Raw data -> Data preprocessing -> Prepared data
Apply machine learning alg -> Candidate Model
Supplied machine learning alg.
Deploy choosen model. 
Azure ML Api - Deploy
Recrease model periodcally
App use a model.
Azure ml generate a model 
Each - how likly is customer to churn. 
Data on azure blobs. 

Termonilogy
------------
Training the model - train a model
Using training data
Superised model - the value is present in training data
Labeled data - contain the value tyring to predict.
Unlabled. 


Data (Relational,no sql,)
Training data - just a table of rows and Columns
Data contain the label. 
Columns - Features. 
One of feature is the target value. 
Apply machine learning algorithm . 
Regression. 
Classification. 
Clustering. Find clusters. 
Few styles
----------
Decision Tree
Nural network 

----
Input only 75% into the learning alg. 
Outcome is candidate model
Is the model preditictive? 
Make prediction - Generate target values. prediction values.
Compare generated target value to actual target values.
1) May be try other alg
2) May be need to 
3) Deploy model - Azure ML - build in support to deploy model to azure. 
4) Application must supply the feature. 
5) Model return the predicted target value.