BY AKSHAY TOSHNIWAL
ROADMAP TO
BUILD A
MACHINE
LEARNING
MODEL
A complete guide to build machine learning
system (an area under artificial intelligence)
1 Roadmap to build a machine learning model
UNDERSTAND THE BASICS
There are machine learning libraries available
that will ease your work, but you need to focus
on the basics before you understand them.
Understand the basics of math and stats that
include topics like, linear algebra, calculus,
probability, and statistics.
2 Roadmap to build a machine learning model
DATA PRE-PROCESSING
Real-time/real-world data is dirty & noisy and it
contains a lot of outliers (unexpected data
records) that are not relevant for the ML
modelling.
Cleaning this is important to avoid any
unexpected results and also to structure the
data.
3 Roadmap to build a machine learning model
IDENTIFYING FEATURE
Any given dataset has multiple features and
using these features the machine learning
model is designed.
Two ways to do this, selecting the features
(feature selection) or creating new features
(feature engineering). Relevant feature
identification can make the model more
effective.
4 Roadmap to build a machine learning model
DIVIDE THE DATASET
The entire dataset needs to be divided into
majorly three chunks: training data, validation
data, and testing data
Training data is used to build (fit) the model,
validation data is used to test the model with
known values, and testing data is used to test
the model with unknown values.
5 Roadmap to build a machine learning model
IDENTIFYING ALGORITHM
Based on the dataset, an effective selection of
an algorithm is very important. There are three
types: Regression, Classification, and
Clustering.
Regression and Classification are supervised
learning algorithms (have known labels), while
Clustering is an unsupervised learning
algorithm (no known labels).
6 Roadmap to build a machine learning model
BUILDING THE MODEL
Using the training data, the model is built and
it is further tested with the validation data.
Fine-tuning the model is performed (if need
be) and multiple iterations are performed to
check whether the outcomes are updating or
not.
7 Roadmap to build a machine learning model
EVALUATING THE MODEL
The most important step in building any ML
model is to evaluate it and check for errors.
This step is mandatory to check whether the
algorithm that is applied is effective/accurate
or not and it is done by calculating different
scores, error terms, and other metrics.
8 Roadmap to build a machine learning model
USE TESTING DATA
To further cross-check the model, use testing
data (unknown data values) to generate the
outcomes.
Again evaluate the model and generate
different metrics. If need be fine-tune the
model and reiterate (Go to Step 6).
9 Roadmap to build a machine learning model
BUILD API AND GUI
Once the modelling is completed and the
testing is successful, build application
programming interfaces (APIs).
Connect these APIs with a graphical user
interface (GUI) for the end-user to experience
the machine learning system.
10 Roadmap to build a machine learning model
DEPLOY THE MODEL
Post building the user interface and
connecting all the APIs, the system is ready to
be deployed on a cloud server
A few of the cloud service providers are
Microsoft Azure, Google Cloud Platform, and
Amazon Web Server (AWS). There are others as
well, but these three provide great tools to
manage ML models.
I hope you found
Follow me for more such insightful posts
this roadmap
insightful and useful
Follow Me: Support Me:
Roadmap by Akshay Toshniwal
