Jason Fang - Portfolio

All
Machine Learning
Deep Learning
Statistical Analysis
Visualization
SQL
NLP
AWS

Tableau Showcase

Here are some of my Tableau examples. These worksheets and dashboards were published on Tableau Public and were embedded on my website. You can see some interactive worksheets and dashboards here and try them by yourself.

Click the "Details" button to see my visuals.

Details

Tableau Showcase

Machine Learning, Visualization

Website SQL Analysis

Using SQL to analyze website performance is a daily routine for data scientists. It's not necessary to always use machine learning and deep learning models to create predictive models. Solving business problems is the top task for us rather than creating models. In this project, I use an e-commerce website dataset to analyze some website problems such as traffic source, bounce rate and so on. Even though I'd like to create some predictive models in real life, I only use SQL in this project.

Software used: SQL
Click the "Details" button for more details.

Details

Website SQL Analysis

SQL

Fintech App

This is a real fintech company that wants to provide its customers with a paid mobile app subscription that will allow them to track all of their finances in one place. To attract customers, the company releases a free version of their app with some main features unlocked. The purpose of this project is to identify which users will most likely not enroll in the paid product so that additional offers can be given to them. I used several models and finally got 100% accuracy. This is just an example of how I find out the company's potential customers. The techniques I used here could be applied in other companies as well.

100% Accuracy in Testing Dataset

In real life, things are more complicated, so I also have some last thoughts in the end.

Software used: Python

Packages used: Sklearn, Tensorflow, Keras, Seaborn

Click the "Details" button for more details.

Details

Fintech App

Machine Learning, Deep Learning

Time Series ARIMA Models

Time series forecasting is a crucial topic for data scientists. Autoregressive integrated moving average (ARIMA) is one of the famous linear statistical models for time series forecasting. It usually outperforms machine learning and deep learning models for one-step forecasting on univariate datasets. There are different variations of ARIMA. In this project, I use one type of ARIMA model which is SARIMAX, Seasonal Auto-Regressive Integrated Moving Average with eXogenous factors. A restaurant visitors dataset has been used in this project.

Software used: Python
Packages used: statsmodel, pmdarima
Click the "Details" button for more details.

Details

Time Series ARIMA Models

Statistical Analysis

Spark Spam Filter

Data scientists deal with big data by using Spark which is very powerful and user-friendly. Spark 2.0 has shifted towards a dataframe syntax which is just like pandas and Spark's MLlib library is similar to sklearn library. You also can write SQL queries by using Spark. Since I've had many machine learning and SQL projects, I used Spark MLlib to create a spam filter in this Natural Language Processing project.

Software used: VirtualBox, Ubuntu, Jupyter Notebook
Packages used: Pyspark, Spark MLlib
Click the "Details" button for more details.

Details

Spark Spam Filter

Machine Learning, NLP

A/B Testing

A/B testing is a very popular topic among companies. As a data scientist, there are several things that I need to check before I conduct A/B testing. The first half of this project is the workflow I'd go over every time. There's also an example of A/B testing in the second half.

If an A/B testing experiment has multiple treatment groups, the cost of testing each treatment group and the probability of false positives will increase. When there are multiple treatments groups, I usually use two types of reinforcement learning, Upper Confidence Bound (UCB) Algorithm and Thompson Sampling to quickly find out the best treatment group by using the minimum amount of time. You can see how I use reinforcement learning to conduct A/B testing in A/B Testing Reinforcement Learning Project

Software used: Tableau

Click the "Details" button for more details.

Details

A/B Testing

Statistical Analysis, Visualization

Self-Organizing Maps

Self-Organizing Map is a type of unsupervised deep learning model. The purpose of using SOMs is to reduce the dimension of our dataset and create visuals. In this project, I combined SOMs and the Artificial Neural Network to create a model for fraud detection. We don't want to see that our customers are cheated, and then we perform supervised learning to protect them in the future. Fraud has to be stopped before it happens.

Software used: Python, Pycharm
Packages used: Tensorflow, Keras, Sklearn, MiniSom, pylab,
Click the "Details" button for more details.

Details

Self-Organizing Maps

Deep Learning

Fraud Detection

This is a preprocessed dataset. There are 28 features not interpretable because the personal information of our customers has been transformed and kept anonymous. The "Class" feature is whether customers have been defrauded or not, 0 = no, 1 = yes. I used undersampling and oversampling in this case because this dataset contains more than 280,000 observations, but only about 400 people have been defrauded, which just accounts for 0.14% of the dataset.

Software used: Python

Packages used: Sklearn, Seaborn, Tensorflow, Keras, imblearn

Click the "Details" button for more details.

Details

Fraud Detection

Machine Learning, Deep Learning

Yammer Case

I used AWS Redshift to investigate a drop in user engagement for Yammer. Yammer is an enterprise social networking service used for private communication within organizations. This is a relatively small dataset, with more than 34,000 rows of records and 5 tables, but I used this demo to show you my thinking process when I face a business problem.

Software used: AWS Redshift, AWS S3, AWS IAM

Click the "Details" button for more details.

Details

Yammer Case

SQL, AWS

Self-Driving Car

Since I’m very interested in the self-driving car industry, I used Python to create a Convolution Neural Networks (CNN) model to simulate a self-driving car. It’s very similar to the ongoing self-driving experiments conducted by Tesla, Google, Uber, and so on. My process is simpler, but the idea is the same.

1. Used a self-driving simulator provided by Udacity to generate the data.

2. Balanced the data and split the data into the training and validation data set.

3. Preprocessed the data and used OpenCV.

4. Used Convolution neural network CNN to create a model.

5. Tested the model by using the code provided by Udacity.

Software used: Python, Pycharm, Git, Udacity self-driving simulator

Packages used: tensorflow, keras, sklearn, cv2, pandas, numpy, os, matplotlib, imgaug

These are simplified steps. Click the "Details" button for more details or the "Code" button to see the code of this project.

Code Details

Self-Driving Car

Deep Learning

Football Pose Estimation

I did this project to help Tencent to create some educational football tutorials. As a data scientist, I’ve applied one of the most useful tools, OpenCV, in the process. By using OpenCV, people can track the coach’s pose and hands positions, so that they can throw or catch the football correctly when they start learning how to play football. his one-minute-long video is a demo of my work.

Software used: Python, Pycharm, Git, Adobe Premiere

Packages used: numpy, cv2, mediapipe, uuid, os, time

These are simplified steps. Click the "Details" button for more details or the "Code" button to see the code of this project.

Code

Football Pose Estimation

Computer Vision

Matplotlib & Seaborn Showcase

Here are some of my Matplotlib and Seaborn examples. I usually use Tableau during the data exploration and reporting stages. However, Matplotlib and Seaborn are also very useful when I write Python code. I use Matplotlib and Seaborn to generate kernel density estimation plots, pie charts, box plots, and so on. While only generating correct plots is not enough. I do care about the style and color of my plots.

Click the "Details" button to see my visuals.

Details

Matplotlib & Seaborn Showcase

Machine Learning, Visualization

Football Pose Estimation

This game is mainly designed by using object-oriented programming and pygame package. I created a lot of classes to define what the player spaceship and enemies spaceships should move in the game. What should the game react when the player controls his spaceship.

Software used: Python, Pycharm, Git, Adobe Premiere

Packages used: pygame, os, random, time

Click the "Details" button for more details or the "Code" button to see the code of this project.

Code Details

Space Invaders Game

Object-Oriented Programming

SQL & RedShift

SQL is commonly used by data analysts and data scientists. I had used different types of SQL, such as SAS proc SQL, PopSql, MySQL Workbench, MS SQL. In the end, I decided to use Amazon Redshift. First, I don't need to install any SQL software on my laptop or computer. It can run a large database on the AWS server rather than on my antiquated laptop. Second, I'm very familiar with AWS services, and I use AWS S3 a lot and upload almost all of my work to S3. Third, Redshift is very powerful and user-friendly. I could generate charts directly from my query in Redshift. Also, I can see my queries' history and share them with my colleagues.

Software used: AWS Redshift, AWS S3, AWS IAM

Click the "Details" button for more details.

Details-->

SQL & RedShift

SQL, AWS

A/B Testing Reinforcement Learning

A/B testing is a very popular topic among companies. You can find my workflow in A/B Testing Project, which contains an A/B testing example by using Tableau. However, real-life problems are usually more complicated. If an A/B testing experiment has multiple treatment groups, the cost of testing each treatment group and the probability of false positives will increase. When there are multiple treatments groups, I usually use two types of reinforcement learning, Upper Confidence Bound (UCB) Algorithm and Thompson Sampling to quickly find out the best treatment group by using the minimum amount of time.

Software used: Python

Click the "Details" button for more details.

Details

A/B Testing 2

Statistical Analysis

Breast Cancer Classification

People create tons of models every year, but 85% of them are useless. As a data scientist, I want to create values for my company and the world. In practice, good models can save lives and bad models can cost lives.

This is an example of how I use classification models to identify malignant and benign for breast cancer. Usually, a 90% accuracy model is good enough, but it's not good enough for identifying cancer. Even though this is just an example and no one would use my models, I still treat this example seriously. That's what data scientists are about to do in the future, changing the world.

Software used: Python

Packages used: sklearn, seaborn, GridSearchCV, numpy, pandas

Click the "Details" button for more details.

Details

Breast Cancer Classification

Machine Learning

Echo Look

In this project, I used an image dataset that contains 58,000 images of apparel, such as T-shirts, trousers, pullovers, dresses, and so on. The images are in greyscale, and they are just 28*28 pixels. By using the convolutional neural network, I'm able to identify different types of apparel. It can be implemented in the real world and Echo Look is an example of it. This project just shows how I use CNN to classify apparel.

Software used: Python

Packages used: Sklearn, Tensorflow, Keras, Seaborn

Click the "Details" button for more details.

Details

Echo Look

Machine Learning, Deep Learning

Cat vs Dog Classifier

Classifying a cat or dog image sounds very easy, but nobody could have a 100% accuracy model nowadays. There is a famous meme about AI, which is that AI identifies a cat as a dog. In this project, I ended up with an 88% accuracy model in my testing dataset. The model is just example of how I created a Convolutional Neural Network model.

Software used: Python, Pycharm
Packages used: Tensorflow, Keras
Click the "Details" button for more details.

Details

Cat vs Dog Classifier

Deep Learning

ETL Pipelines

Building ETL pipelines should be data engineers' jobs. While as a data scientist, it's also necessary to know the fundamental ideas of data pipelines. In this project, I choose a public dataset from Deutsche Börse Public Dataset. By using AWS S3 and python, I'm able to extract the data from the original AWS S3 bucket and transform the data, then load the data to my AWS S3 bucket.

Software used: Python, AWS S3
Click the "Details" button for more details.

Details

ETL Pipelines

Machine Learning

AWS QuickSight

I used to use Tableau during the data exploitation and reporting stages until I started using AWS QuickSight. AWS QuickSight is composed of Quick and Sight. After using it, I feel QuickSight is too quick and too insightful. Due to my Tableau background, I found out that QuickSight is very similar to Tableau and it also has some features that Tableau doesn't have. In this project, I'll show you my experience of using AWS QuickSight.

Software used: AWS QuickSight
Click the "Details" button for more details.

Details

AWS QuickSight

Machine Learning, Visualization, AWS

Stock Price Prediction

In 2012, some scholars from Stanford published an article saying that they could use the Recurrent Neural Networks (RNN) to predict the stock price. I knew that RNN models could make powerful time-series predictions because of long short-term memory (LSTM). While I didn't expect that it could be used to predict the stock price, so I tried it myself, and I also got a very accurate model.

Software used: Python, Pycharm
Packages used: Tensorflow, Keras
Click the "Details" button for more details.

Details

Stock Price Prediction

Deep Learning

Loan E-Signing

In this project, I worked for a fintech company that wanted to minimize the churn rate for its app. Mainly focusing on the app features, I could help the company improve the app to keep its customers. To create the best model, I applied feature selection, data balancing, k-fold cross-validation and some other techniques. It is very hard to predict that whether people would churn or not because the churn rate is not balanced, only about 40% of people churned. In the end, I created a model with a 64% accuracy.

Software used: Python

Packages used: Sklearn, Seaborn

Click the "Details" button for more details.

Details

Loan E-Signing

Machine Learning

Churn Rate Analysis

Software used: Python

Packages used: Sklearn, Seaborn

Click the "Details" button for more details.

Details

Churn Rate Analysis

Machine Learning

IMDB Rating Analysis

Natural Language Processing (NLP) is everywhere in our life. There are many types of NLP, such as classical NLP, deep learning NLP, seq2seq, and so on. This project is the first part of my NLP project which covers a classical NLP model, bag-of-words, for sentiment analysis.

Software used: Python

Packages used: nltk, sklearn, re

Click the "Details" button for more details.

Details

Natural Language Processing

Machine Learning, NLP

IMDB Rating Analysis

Natural Language Processing is playing an important role in the machine learning field. I’d like to share a powerful NLP tool which is provided by IBM, IBM Watson. Watson is a question-answering computer system capable of answering questions posed in natural language, developed in IBM's DeepQA project by a research team led by principal investigator David Ferrucci.

It can easily be used in Python. You can just install ibm_watson and import the module that you want to use. There are 4 modules that I use very often, Text to Speech, Speech to Text, Translation, and Tone Analysis. Text to Speech is very accurate because the computer can read every single word. While Speech to Text accuracy is not that high, even you use the audio that is converted by Text to Speech. Translation provides a very decent service but not as good as google translation. Tone Analysis could help you analyze your sentence’s tone. It’s been used in interpreting reviews and Twitter content. This powerful tool can help you find out whether the comment is positive or negative.

Software used: IBM Watson, Google Colaboratory

I have created a demo document in Google Colaboratory. Click the "Code" button to see the code of this project.

Code

IBM Watson

Machine Learning, NLP

Boltzmann Machines

Boltzmann Machine is an unsupervised deep learning model. It has no direction, and it's not a deterministic model. In deterministic models, the output of the model is fully determined by the parameter values and the initial conditions initial conditions. Boltzmann Machine is a stochastic model which possesses some inherent randomness. The same set of parameter values and initial conditions will lead to an ensemble of different outputs. A movie dataset has been used which has more than 24,000,000 ratings and 670,000 tags applied to 40,000 movies by 260,000 users.

Click the "Details" button for more details.

Details

Boltzmann Machines

Machine Learning

Linear Regression Algorithm Implementation

This is the process of how I implemented a linear regression algorithm from scratch, how I created the coefficient function, and how I created the R-square function. You may wonder why I didn't just import the algorithm from scikit-learn directly. The main reason is that as a data scientist, you not only need to know how to implement a model, but also you have to know the algorithm inside of your model so that you can create a better model.

Linear regression is the simplest machine learning algorithm. It’s so easy that most people don't even realize it’s a machine learning algorithm. It’s just Y = mX+b and you only need to write a few lines of code to implement a linear regression model to your dataset. By implementing a machine learning model, you could easily get a 90% or 95% accuracy by just using the default values. However, sometimes a 99.999% model is not accurate enough like the self-driving car. As a data scientist, I really need to focus on the root cause of an algorithm, so that I can push the limits of my models.

Software used: Python, Pycharm, Git

Packages used: statistics, numpy, random, matplotlib

Click the "Details" button for more details or the "Code" button to see the code of this project.

Code Details

LR Algorithm Implementation

Machine Learning

Linear Regression Algorithm Implementation

This is the process of how I implemented a K-Nearest Neighbors algorithm from scratch. K-Nearest Neighbors is one of the most popular machine learning models. To understand how it really works, I implemented the algorithm from scratch. In the end, I compared my KNN model to sklearn’s KNN model. The accuracies are almost the same.

Software used: Python, Pycharm, Git

Packages used: pandas, numpy, random, matplotlib, sklearn, warnings

Click the "Details" button for more details or the "Code" button to see the code of this project.

Code Details

KNN Algorithm implementation

Machine Learning

IMDB Rating Analysis

I used some linear regression and machine learning models to analyze an IMDB dataset which contains 14166 movies and 72 variables. These variables are audience ratings and demographic information.

I first created a linear regression model to find out which variables are crucial to the movie box office.

Then I used the Forward Stepwise Regression model and Full Model Search model to delete some variables to decrease over-fitting.

Finally, I used the decision tree and random forest model to find out which age group has the greatest influence on the movie rating by using the Variable Important Plot.

Software used: R Studio

Packages used: Linear Regression, Forward Stepwise Regression, Full Model Search, Decision Tree, Random Forest

Click the "Code" button to see the code of this project.

Code

IMDB Rating Analysis

Machine Learning

Web Development HTML

My personal website developed by using HTML, CSS and JavaScript. As a programmer and magician, I need two separate websites for two kinds of audiences, but I also need to integrate these two websites together. It's a simple static website, but I developed this website by myself. A lot of skills are required to build this website.

Software used: HTML, CSS, JavaScript, Intellij IDEA, Git, Adobe Photoshop, Adobe Premiere, AWS S3, AWS Route 53, AWS CloudFront

Click the "Details" button for more details or click the "Code" button to see the code of this project

Code Details

Web Development HTML

Marketing Analytics, Digital Marketing

Web Development WordPress

OmniFun is a kid’s playground located in Maryland, Gaithersburg. Because other managers don’t know HTML, I chose to use WordPress to build this website. This website is designed to show parents the pictures, pricing, and policies. I also implemented Google Analytics which can track online viewers' information.

Software Used: WordPress, Google Analytics

Click the "Visit" button to visit the website or click the "Details‘ button for more details.

Visit Details

Web Development WordPress

Marketing Analytics, Digital Marketing

Graphic Design

Digital Marketing, Graphic Design

MailChimp Newsletter

Mailchimp is an American marketing automation platform and email marketing service. Since I took over the email marketing, our opening rate has increased from 10% to almost 30%. I usually design a poster at first and then implement some pictures and call-to-action button.

Software used: MailChimp, Adobe Photoshop

MailChimp Newsletter

Marketing Analytics, Digital Marketing

Social Media Management

I had managed some company social media accounts on different platforms including Facebook, Instagram, YouTube and so on. We have reached more than 20k people on Facebook.

Software used: Facebook, Instagram, YouTube

Click the "Visit" button to visit the social media

Visit

Social Media Management

Digital Marketing, Social Media

Tableau Showcase

Website SQL Analysis

Fintech App

Time Series ARIMA Models

Spark Spam Filter

A/B Testing

Self-Organizing Maps

Fraud Detection

Yammer Case

Self-Driving Car

Football Pose Estimation

Matplotlib & Seaborn Showcase

Space Invaders Game

SQL & RedShift

A/B Testing 2

Breast Cancer Classification

Echo Look

Cat vs Dog Classifier

ETL Pipelines

AWS QuickSight

Stock Price Prediction

Loan E-Signing

Churn Rate Analysis

Natural Language Processing

IBM Watson

Boltzmann Machines

LR Algorithm Implementation

KNN Algorithm implementation

IMDB Rating Analysis

Web Development HTML

Web Development WordPress

Graphic Design

MailChimp Newsletter

Social Media Management

Let's work together on your next project