Client website External

Client Social Media Dataset

Pixel data Sources Uploads

Data Extraction and Data Type Selection

Analytics Frameworks
transformation from
source
Google Analytics Storage
(Script or ETL
Facebook Analytics
Service)
Classification
Time Document Images and
and
Series and Text Videos
Regression
Dynamo DB Cassandra MySQL Redis MongoDB

Fast, Scalable Highly Scalable Secure Fast & Highly Secure Very Fast
High Uptime Fault Tolerant Moderately Supports Graph Highly Scalable Histogram
Data Visualisation
Low Maintenance High Speed Fast Higher Complexity Better for front end using
MatPlotLib and Seaborn
Scatter Plot

Data Preparation

Data
Transformation Data Cleaning

Normalization
using Min-Max or Missing Noise Outlier
Z-score Values Cleaning Detection

Dxi Calculation Replace

Ignore the
the Binning
tuple DBSCAN Normalization
value method Techiniques
Feature Selection

with
with majority
Unsupervised Supervised with zero
mean/median repeated
values

AntiLog Left Right Log

Wrapper Data Skew?
Intrinsic Filter method Operations Operation
Method
No

Recursive Convert to normal

Check for
Trees Feature Stats Feature Importance distribution with
Outlier
Elimination mean 0 and sigma 'n'

Imbalance Data

Upsampling Downsampling SMOT

Data Reduction
(Not Compulsory)
(only for unsupervised)

PCA Wavelet Transform

Common Features Selected

ML/DL Algorithms

Ensemble Learning
Probability Prediction Clustering
Classification Time Series
KNN (Classification
Bagging
& Regression) Naive Bayes K-Means
Boosting ElmanCart Arima Forecast
RFM Random Forest Heirarchical clustering
Stacking SimpleCart LSTM
Bayesian Learning K-mediot
Random Forest

Conversion to Real Values

Front-End
PreProcessing

Manual and
Automated Testing

Finish