Zihan Zeng

Detail-focused Data Analyst with knowledge in data warehousing, process validation and business needs analysis. Proven to understand customer requirements and translate into actionable project plans. Dedicated and hard-working with passion for Big Data. Seasoned collaborator experienced in meeting needs, improving processes and exceeding requirements in team environments. Diligent worker with strong communication and task prioritization skills.

Unicom Customer Portrait in Precision Marketing                            Jan. 2022-Mar.2022

Position: group leader

· Used an open data sets on SODA website, with 280,000 user-related information; employed exploratory data analysis method to analyze customer characteristics, and K-means algorithm to cluster users with different characteristics to form customer portraits; 

Recommended suitable package types to users and proposed feasible precise marketing strategy based on the portrait characteristics.

Data-driven Forecasting of Credit Default                                        Sept. 2021-Dec.2021

Position: group leader

· Imported a publicized data set on Kaggle and applied a procedure of Exploratory Data Analysis;

Normalized the datasets and apply deal with missing value: drop some data with large proportion of missing value and use median to fill the rest of missing values. Employed Logistic Regression, Random Forrest and LightGBM to devise prediction model; compared the efficiency of different algorithms.

TSA Analysis of the Target Stocks                                                     Mar. 2021-May.2021

Position: group member

Acquired time series data of bond yield from Wind and perform regular data cleaning and normalization;

Identified the presence of AR and MA components in the residuals by utilizing Matlab functions; devised the best fit ARIMA model by calculating the p, d, and q components;

Applied devised model to predict the future trends of the value, validated the prediction with the real time indictors, and suggested the approaches to improve the current modeling strategies.