遊戲數量眾多，屬性、標籤、遊戲種類、熱門程度等變數多而混雜。從資料處理上來說，是屬於高維度而稀疏的資料。視覺化遊戲圖能一次性呈現數千款遊戲，並包含遊戲名稱、遊戲群落、熱門程度等資訊，同時以遊戲距離反映近似程度。以此為基礎可以幫助使用者尋找遊戲並探索其關聯，如同樣為動作遊戲，包含解謎要素與否，這樣的差異如何表現在遊戲距離上。借助視覺化的功能，可以觀察到遊戲的不同標籤或屬性對遊戲相似度的影響，並直觀以距離體現。

遊戲圖中以圖標尺寸反映遊戲熱門程度，藉此得以快速找到自己熟悉的遊戲，並以此為基礎找到相近遊戲，此外也能用以發掘新遊戲。相比起Steam的搜尋列表，遊戲圖能夠以清楚的圖標反映熱門程度，並藉由瀏覽遊戲整體建立對多樣遊戲的認識。同時以聚落的形式表現遊戲，形成的遊戲群不只利於瀏覽，也使挑選近似遊戲更加快速。使用者藉由視覺化的遊戲圖，除了能夠全面地看到遊戲的整體資訊，還能根據偏好遊戲所在的聚落，進行拉近聚焦檢視以尋找可能的目標遊戲。如以遊戲《人中之龍》為例，顯示相關遊戲如《Azur lane crosswave》、《Evoland2》等。

The sheer volume of games, mixed with various attributes, tags, types, and popularity levels, result in high-dimensional and sparse data from a data processing perspective. A visualized game map can simultaneously present thousands of games, providing information such as game names, clusters, and popularity. It also indicates the similarity between games through the distance displayed on the map. This basis assists users in locating games and exploring their relationships.

For example, it can be observed how differences, such as the presence or absence of puzzle elements in action games, are reflected in the game distance. With the help of visualization, the impact of different game tags or attributes on game similarity can be observed and intuitively represented by distance.

In the game map, the bubbles reflect the popularity of the games, enabling users to locate familiar games quickly, find similar games, and discover new ones. Compared to Steam's search list, the game map can display popularity with distinct icons and allows users to understand various games within a holistic view. Representing games in clusters facilitates browsing games and speeds up selecting similar games. Users can access comprehensive game information through the visualized game map and focus on potential games by targeting the clusters where their preferred games are located. For instance, using the game "Yakuza" as a reference, related games such as "Azur Lane Crosswave" and "Evoland 2" are displayed.

1 緒論 - 1 -
1.1 研究動機 - 1 -
1.2 章節說明 - 2 -
2 資料準備 - 3 -
2.1 資料背景 - 3 -
2.2 資料蒐集 - 3 -
2.3 資料整理 - 5 -
3 研究方法 - 7 -
3.1 數值變數 - 7 -
3.2 類別變數 - 9 -
3.3 LDA features - 10 -
3.4 MDS and t-SNE - 11 -
4 遊戲圖 - 13 -
4.1 尋找遊戲《人中之龍》 - 13 -
4.2 局部遊戲檢視與瀏覽 - 15 -
4.3 發掘不同風格遊戲 - 18 -
4.4 視覺化與遊戲群落 - 21 -
4.5 討論 - 23 -
5 結論 - 25 -

[1] Blei, D. M., Ng, A. Y., and Jordan, M. I. (2003). Latent Dirichlet Allocation. Journal of Machine Learning Research, 3(Jan), 993-1022.

[2] Härdle, W., Simar, L., Härdle, W., and Simar, L. (2003). Mult-idimensional scaling. Applied Multivariate Statistical Analysis, 373-392.

[3] Davis, N. (2019). Steam Data Collection.
[URL: https://nik-davis.github.io/posts/2019/steam-data-collection/]

[4] Li, X., and Zhang, B. (2020, January). A preliminary network analysis on steam game tags: another way of understanding game genres. In Proceedings of the 23rd International Conference on Academic Mindtrek , 65-73.

[5] Lin, J.-H. (2023). Visualization of Steam Games.
[URL: https://facemask01.github.io/thesis.io/CH4_4]

[6] Lin, J.-H. (2023). Steam raw data.
[URL: https://facemask01.github.io/thesis.io/STEAM RAW DATA-Copy1]

[7] Van der Maaten, L., and Hinton, G. (2008). Visualizing Data Using t-SNE. Journal of Machine Learning Research, 9