Qijia He

The network of relationships between people is a fascinating subject. How many friends of my friends are connected to Bill Gates? Back in the 1960s, renowned American social psychologist Stanley Milgram conducted a famous letter-passing experiment, which led to the well-known Six Degrees of Separation theory and the small world phenomenon. The rise of the internet further spurred a wave of research into social networks. In 2000, Jon Kleinberg proposed a mathematical model for social networks. For this project, I have made some modifications to the long-range connections in Kleinberg's model to explore how changes in assumptions can affect the network. Specifically, I investigate which power-law distribution can optimize the network under the modified conditions.

Our report is available here. Github page: https://github.com/Heqijia/STA5001

The fluctuations of financial markets are always thrilling. Analyzing a company's performance and predicting stock trends have long been the tasks of banks and investment analysts. In this project, we analyzed companies' positions and operating conditions within their respective industries based on their specific financial statements. We also made rough predictions about stock price movements to provide investment recommendations for investors within specific sectors. Our data was sourced from baostock.com, focusing on the financial statements of companies for the fourth quarter of 2023. After categorizing and rating the companies, we offered stock-picking advice for the industries. Among the three industries we analyzed (machinery, chemical, and biomedical), our recommended investment portfolios significantly outperformed the average stock price fluctuations in these sectors, using the opening prices on April 1, 2024, as a reference.

Our report is avaliable here. Github page: https://github.com/Heqijia/MA304-proj

Taking the subway is always frustrating; the constant opening and closing of doors and the frequent stops and starts of the train waste a lot of commuters' time. Improving commuting efficiency with existing resources is a significant aspect of building smart cities. Based on the passenger flow data of Shenzhen Metro Line 5 on August 31, 2018, and without considering the regularity of station stops, we optimized the train stops during the morning peak. In this project, we designed three different greedy algorithms and further optimized them with local search. We found that during peak congestion periods, adopting a flexible skipping-stop strategy can save passengers 10% to 15% of their average commuting time.

Click these 2 links to see our improvements on metro data: 1. Station Passenger Flow, 2. Metro Schedule.

Our report is avaliable here. Github page: https://github.com/Heqijia/STA326-Project

This is a visualization dashboard built with D3 to display transaction data in the US secondhand car market. The dataset, containing about 500,000 transactions, was directly downloaded from Kaggle. The dashboard includes several components:

• A pie chart showing the distribution of car types for a selected brand.
• A word cloud presenting the brands and their proportions in the secondhand market.
• A bar chart showing the top 7 best-selling cars of the selected brand.
• A parallel coordinate display detailing the specifications of the cars.
• A prediction model that estimates a car's price based on its model, year, and odometer. The result is returned and displayed on the scatter plot. This prediction model is simply implemented using KNN.

Click here to play the dashboard!

Our report is avaliable here. Github page: https://github.com/Heqijia/CSE332_lab5

The face recognition algorithm has been continuously refined and widely applied in contemporary society. In this course project, we encapsulated the Deepface model and created a user interface for a face recognition system using Python's PyQt5. and did performance analysis based on our own dataset. The system comprises five main functions:

1. Face attribute analysis: Input an image and get predictions for age, ethnicity, gender, and emotion.
2. Video Face detector: Detect and track faces in a video input.
3. Image Face detector: Detect faces in a still image.
4. Find matching Faces: Built a local face database comprising photos of three project members and various online personalities. Compare newly input images with the database and return whether they match any database photos and who they might be.
5. Face Verification: Input two photos separately and determine whether the faces in the photos belong to the same person.

For more details, you can view our report or github Repository:

Our report is avaliable here. Github page: https://github.com/Fae42/cv-final-project

Bilibili has gradually matured over the past few years in China (similar to YouTube), becoming a highly concentrated cultural community and video platform for the younger generation in China. As the number of Bilibili users and vloggers has increased, Bilibili has come to reflect the value orientations of China's younger generation to some extent. Consequently, the importance of big data research on Bilibili has become increasingly evident.

By analyzing the behavior of Bilibili's vloggers, we can gain a clearer understanding of the psychological needs and interests of today's youth, as well as the relationship between certain behaviors of vloggers and their number of fans. This allows us to better grasp the current mainstream values and provide relevant suggestions for both new and experienced vloggers on creating videos and attracting fans.

Dataset:

• A CSV file containing basic information of 5k vloggers in 2019, from Prof. Yifang Ma
• A tracking data based on the ID of the vloggers (2022), collected by Google Web Scraper

Our report is avaliable here. (In Chinese)

This project aims to develop a solver for a number puzzle game. Given any initial configuration, the goal is to determine if there is a solution. In this puzzle, 0 represents an empty space, yellow blocks represent large blocks (which can be 2x1, 1x2, or 2x2), and blue blocks represent small blocks (which are 1x1). For example, the final solution should look like [[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], [13, 14, 15, 16], [17, 18, 0, 0]]. In the searching process, only the blocks adjacent to the 0 block can be moved.

We implemented three solvers: Solver, Solver1, and Solver2. The Solver uses a self-designed greedy algorithm, Solver1 uses pure BFS (Breadth-First Search), and Solver2 uses pure DFS (Depth-First Search). The greedy algorithm in Solver prioritizes the move of blocks based on the distance between their current dashboard structure and the target dashboard structure, giving higher priority to the move of blocks and structures that will lead the board closer to its final structure. A Min Priority Queue (MinPQ) is used to manage the board's search structure. When the size of MinPQ exceeds 500,000, we retain only the best 10,000 points. This implementation performs significantly better than Solver1 and Solver2, allowing Solver to handle boards as large as 6x6 in a limited amount of time, while the other two can only handle up to 3x3 boards. For detailed information on the algorithms, please refer to [...] (Written in Chinese)

Once the search process is complete, a final dashboard will be generated, displaying the steps taken to arrive at the solution.

Our report is avaliable here. Github page: https://github.com/Heqijia/DSBB-project

Xiaoyang Wang is a member of the Shenzhen Dive4love club, a newly minted diving instructor, coral reef conservationist and educator on the Dapeng Peninsula, Shenzhen. Despite being a latecomer to this field, Wang has an immense passion for nature. In this interview, we mainly discussed why Wang chose diving and coral conservation as his career, the current state and challenges of coral reef conservation, people's understanding and awareness of nature conservation, the role of education, and the social conflicts between environmental protection and development. The detail content can be seen in this report.

Our report is avaliable here.

In this project, a chess game system is developed using Java (the interface is in Chinese). All rules of chess are implemented in this system. The interface is created using javax.swing. Click hello.exe to start the game! Some supplementary functions include:

• Hint for the next move, displayed as green dots.
• Countdown: each player has 45 seconds to make a decision; otherwise, the turn automatically switches to the opponent.
• Music and various background settings.
• Game saving: this feature generates a text file containing the basic information of the chessboard, including whose turn it is.
• Undo a move: multiple undos are allowed.

No report is abaliable for this project. Github page: https://github.com/Heqijia/IntroCP-Project