SUMMARY

• More than 5+ years of experience working on the Data science tools and techniques such as Risk Modeling Fraud detection and A/B Testing focused on Credit card fraud detection modeling algorithm, marketing company conversion rate prediction and E - commerce company products A/B testing with Python programming, Java, Django, Flask, database design and agile methodologies.
• Have a good experience working with deep learning neural networks. Proficient with Statistics, Data Analytics, Machine Learning using Python.
• Assisting business by being able to deliver a machine learning project from beginning to end, aggregating and exploring data, building and validating predictive models and deploying completed models to deliver business impacts to the company.
• Experienced in SQL programming and performed ETL operations on NoSQL/MYSQL database.
• Used Apache Spark for data processing on large data sets using Hadoop and SparkML for machine learning pipelines.
• Performed data cleaning, features engineering using Pandas and Numpy packages in Python and build machine learning models using Scikit-Learning package in Python.
• Used Python and R for Exploratory Data Analysis, A/B testing, pySpark, H2O, AWS, ANOVA test to compare and identify the effectiveness of creative campaigns.
• Proficient in Matplotlib and Seaborn data visualization tools to analyze and obtain insights into large datasets.
• Experienced with Machine Learning techniques such as Linear Regression, Logistic Regression, Naïve Bayes, Random Forest, KNN, SVM, Neural Network, Lasso Regression, k-means.
• Experienced with Deep Learning techniques such as CNNs, RNNs, LSTMs and Statistical Modellings in Forecasting / Predictive Analysis, Segmentation methodologies, Regression based models, Hypothesis testing, Factor analysis / PCA.
• Created a Handler function in Python using AWS Lambda that invoke when service executed.

TECHNICAL SKILLS

Programming Languages: Python, R, SQL, Java, C++, C

Databases: MySQL, PostgreSQL, Oracle, SQLite, MongoDB

Frameworks: Django, Flask, Tensorflow, Keras

AWS Services: EC2, S3, Auto Scaling, EBS, IAM, AMI, Lambda functions

Libraries: Numpy, Pandas, SciPy, Seaborn, Matplotlib, sklearn, re, PyMongo

PROFESSIONAL EXPERIENCE

Confidential, Chicago IL

Python/Data Scientist

Responsibilities:

• Involved in retrieving “Fraud data”, “IP address to country” data from database and developing reports based on the business requirements.
• Generated reports for Teradata and Oracle Database to analyze the customer behavior and plan strategies to improve the response rates of marketing campaigns.
• Exploratory Data Analysis (EDA) with Python packages such as Pandas, Numpy, Matplotlib, Seaborn for the fraud data and adopted feature selections.
• Implemented a Python script to manipulate the corresponding country with “IP address to country” data and merged the generated country data with “Fraud data”.
• Defined the characteristics of fraudulent activities as many different user ids using same device, many different users having the same ip address and unusual week of the year and unusual day of a week can be the indicators for fraudulent activities, proposed feature engineering and figured out the results for the problem assumptions.
• Set up H2O cluster environment, generated training and test sets and deployed the Random Forest model for fraudulent activities classifications.
• Visualized the importance of features and computed precision, recall, f1-score to evaluate the predictions, plotted ROC curve with AUC scores to check the overfitting of the predictions.
• Results discussions based on ROC curve and minimized false positive to reduce misclassifications on non-fraud users classified as fraud users, made decisions and provide suggestions.

Environment: Python (Pandas, Numpy, Matplotlib, Seaborn, sklearn), h2o, ETL, Oracle, MS Office, AWS, Apache Spark.

Confidential, Alameda CA

Python/Data Scientist

Responsibilities:

• Collected “conversion data” data which involves the information of users and the total number of pages visited of the users, as well as whether the web pages are converted or not is labelled.
• Exploratory Data Analysis (EDA) to check the relationship between countries and conversion rate, the relationship between user types and conversion rate, the relationship between website source and conversion rate, the relationship between user ages and conversion rate and the relationship between total pages visited and conversion rate through visualization.
• Set up H2O cluster, prepared training and test data sets, modelled Random Forest model to classify products converted or not, and generated precision, recall f1-score, support of classification report, as well as ROC curve to evaluate the classification.
• Presented the data science project for the data science, data engineering and software development teams.

Environment: Python (Pandas, Numpy, Matplotlib, Seaborn, sklearn), h2o, ETL, MongoDB, MS Office, AWS, Apache Spark.

Confidential, Madison WI

Python/Data Scientist

Responsibilities:

• Collected “test results” dataset which includes features such as the ids of the users, the date and time when the user hit for the first time company webpage, marketing channel of the web, user device, user operating system, whether the user is in the test, the price of the product which users see and whether the user converted.
• Exploratory Data Analysis (EDA) to check the missing values and data types, proposed data processing and cleaning on the dataset and remove the outliers of dataset, preparing the standard A/B testing dataset and concluded the current problems with the data.
• Data Visualization checks the randomness of features for the users who are under testing and are not under testing.
• Data Manipulation to compute the amount percentage of each feature for the users under testing and users are not under testing, and conclude the initial conclusions that lower price, the higher conversion, however, the average revenue for higher price is larger than the lower price.
• Conducted statistical hypothesis testing to test whether the experimental group’s average revenue is equal or less than the control group’s average revenue.
• Presented the project for the data science, software development and data engineering team and provided actionable insights, such as launching special marketing programs targeted on certain products and encourage some rewards for users to promote consumption.

Environment: Python (Pandas, Numpy, Matplotlib, Seaborn, sklearn), Sqlite, AWS, Oracle, Apache Spark, MS Office.

Confidential

Python/Research Scientist

Responsibilities:

• Extracted traffic data across of one and a half years of the certain road segmentation from the company oracle database, which contains traffic flow and traffic speed of each lane of highway, the average traffic speed and total traffic flow of the mainline, the traffic data is aggregated with 5 mins.
• Data Visualization to illustrate the distribution of traffic flow and traffic speed aggregated with one-day time, data transformation with Python to clean and remove the outliers of data, prepared training and test datasets for the predictions.
• Implemented LSTM model with Keras framework and tried different architecture of the model.
• Trained the model with GPU, predicted mainline traffic speed in 5, 10, 15, 30 minutes, and evaluated the prediction results with RMSE and MAPE methods.
• Proposed Continuous Wavelet Transform (cvt) to detect the traffic congestion and compare the predicted traffic congestion starting time with the real traffic congestion starting time, and concluded with 84% accuracy our model could predict the traffic congestion which would occur in 5 minutes, with 72% accuracy our model could predict the traffic congestion which would occur in 30 minutes.