1. Developed enterprise CDP on the AWS cloud platform, designed dynamic thresholds through order data (original), and established user behavior labels to accelerate the risk control process by 60%+.2. Use the values calculated by the dynamic threshold method in CDP to build a model (collaborative filtering + neural network) to predict user profits and label to assist risk control operations.3. Establish and maintain the quantitative team data pipeline program (Airflow), and automatically execute all business programs in the team.4. Apply XGBoost Regression to predict basketball half-game and full-game odds changes to assist risk control in adjusting basketball in-play game odds.5. Apply Langchain, PandasAI, and Ollama to develop a large language model customized consultant, the database data has been successfully accessed through text input, and charts have been generated.6. · Develop and design a method to recommend a set of lv1 label combinations of potential lv2 labels using associated analysis.7. Develop and design a concentrated betting analysis method to observe the concentrated betting behavior of in-play games in various time periods and use it as a reference indicator for adjusting odds.8. Develop and design related bet analysis methods to find out that different players have highly similar betting tendencies (same settlement results and similar odds), which has been marked.9. Develop and design a simulated odds formula to solve the problem that when the Margin is greater than the original odds and the odds are low, the odds after the margin are less than or equal to 1, resulting in the inability to open the market. You can adjust the parameters to obtain different margin formulas.

Developed an unsupervised model combining economic factors to assist experts in analyzing international cross-border trade patterns that reduce 50%+ processing time.1. This model has been packaged as an invention patent No. I720663 "Multi-Regional Trade Analysis System and Method".2. Got 70% goodness of fit by applying this model to calculate the global strategic pattern of Australian wine manufacturer "Accolade Wines Group."3. Applied the data pipeline program (Airflow) to automatically execute model calculation and web crawler updating data.4. Applied amcharts 4 to implement a large number of data visualization to help users understand data intuitively.5. The concept has also been accepted by the incubator in Keqiao, Shaoxing, Zhejiang, China, and was applied to the following two apps. 5-1. Industry Report Automatic Generator that users can get industry insights reports directly simply by inputting a few fields instead of spending time to collect a lot of data. 5-2. Commodity Futures Customs Data Analysis System that users can quickly find out the country market for potential futures price changes from a new view to analyze the global futures.

1. Expanded business coverage by 15% and increased synergy by analyzing each business service's location coverage and overlap.2. Shortened the time to select housing ads for publishing by 60% by developing the ML model (SVM, XGboost, DNN).3. Developed the house recommendation model by combining customer preferences features (eg., near the park, with a balcony).3-1. Designed new types of recommendation methods (different from collaborative filtering and association analysis), validated by confidence intervals.3-2. Help sales representative speculate user's preferred room type, budget, location...etc.4. Developed group CDP on google cloud platform to get a deeper understanding of customers by concatenating customer data through gathering the common key values of customers.4-1. Designed group customer labels to improve sales representatives to contact customers at the right time, to filter competitors, to speculate about customer life circles, etc.4-2. Developed Flask API for each function on CDP, and enabled Google App Engine service to execute the developed API.4-3. Connected Google Computer Engine with Gitlab to execute the CI function.4-4. Connected Google App Engine and Google Computer Engine to Google Cloud SQL to facilitate accessing data.

1. Completed the firm’s Patent Portfolio (7 Fintech patents) to seize market opportunities and protect their own products on various domains including P2P lending, micro-lending, electronic payment, and financial products.2. Assisted the CEO in inter-company Fintech cooperation projects by giving suggestions on data application of cutting-edge technology, e.g., financial robot models and customer risk-rating models that embedded machine learning and genetic algorithms.

1. Provided advertising advice for Farglory Group through association analysis of commonly used advertising media, preferred locations, preferred room types, and budgets.2. Developed an interactive system to help users without a statistical background to conduct data mining to improve the advertising analysis process.3. Honorable Mention / The 18th Construction Engineering and Management Symposium, Taiwan - Research on the Construction Company's Questionnaire Data Exploration System.