日期： 2025/12/20
課程： 國立臺灣師範大學數學系 - 線性迴歸分析 (Linear Regression Analysis)專題： Kaggle Machine Learning & Data Science Survey Analysis
終於在 12/18 順利結束了這學期最硬、但收穫也最豐富的「線性迴歸」期末報告！
這門課不只是推導 Y=Xβ+ϵ 這麼簡單。這次期末專題，我們挑戰了 Kaggle Machine Learning & Data Science Survey 的真實數據集。這跟課本上那些為了教學而設計好的「完美數據」完全不同，真實世界的資料充滿了雜訊、缺失值 (Missing Values) 和各種需要處理的類別變數。
💡 幾個核心的體悟：
- 清洗資料比建模更花時間 以前覺得跑 Regression 就是把數據丟進軟體、看 P-value 和 R2 就好。但面對 Kaggle 這種萬人填寫的問卷數據，我們花了大半的時間在做 Data Cleaning 和 Feature Engineering。如何把「職稱」、「使用語言」量化放進模型？如何處理共線性 (Multicollinearity)？這些才是決勝負的關鍵。
- 數學是骨架，解釋力是靈魂 模型跑出來不難，難的是「解釋」。為什麼這個變數係數是負的？是否違反了迴歸的基本假設 (Assumptions)？在台上的短短幾分鐘，我們要用數學語言去說服台下的老師和同學，證明我們的模型是有統計意義且符合邏輯的。
- 理論與實務的接軌 作為體育與運動科學系的博士班學生，剛開始沒有辦法習慣嚴謹的證明。但透過這次跨領域的專題合作機會讓我學習到，數學工具是如何應用在分析全球數據科學家的薪資結構或工具選擇上。這種將抽象理論「落地」的過程，真的非常有成就感。
感謝隊友們這幾個禮拜的熬夜趕工，也謝謝老師給的指導，讓我們沒淹沒在數據海裡。雖然過程很燒腦，但看到最終的模型能跑出有意義的結果，一切都值得了！
#NTNU #師大數學 #LinearRegression #Kaggle #DataScience #期末報告 #統計分析
On December 18th, I successfully concluded the most demanding yet intellectually enriching presentation of the semester: Linear Regression Analysis.
This course transcended the mere derivation of the equation Y=Xβ+ϵ. For our final project, we undertook the challenge of analyzing the real-world dataset from the Kaggle Machine Learning & Data Science Survey. Unlike the pristine, pedagogical datasets often found in textbooks, real-world data is replete with noise, missing values, and a myriad of categorical variables requiring intricate handling.
💡 Key Takeaways:
1. Data Cleaning is More Time-Consuming Than Modeling I previously labored under the misconception that regression analysis was simply a matter of feeding data into software and checking P-values and R2. However, when confronted with Kaggle’s massive survey data, we dedicated the vast majority of our time to Data Cleaning and Feature Engineering. How does one quantify "job titles" or "programming languages" for a model? How does one mitigate multicollinearity? These were the critical determinants of our success.
2. Mathematics is the Skeleton; Interpretability is the Soul Executing the model is trivial; the true challenge lies in interpretation. Why is a specific coefficient negative? Does the model violate any fundamental regression assumptions? In our brief time on stage, we were tasked with using mathematical language to convince both faculty and peers that our model was not only statistically significant but also logically sound.
3. Bridging Theory and Practice As a Ph.D. candidate in the Department of Physical Education and Sports Sciences, I initially grappled with the rigorous mathematical proofs required by this course. However, this interdisciplinary collaboration illuminated exactly how mathematical tools are applied to analyze global phenomena—such as salary structures or tool selection among data scientists. The process of grounding abstract theory into practical application was profoundly fulfilling.
I extend my deepest gratitude to my teammates for burning the midnight oil these past few weeks, and to our professor for the guidance that kept us afloat in this sea of data. Although the process was mentally taxing, witnessing our final model yield meaningful results made every moment worthwhile.
#NTNU #Mathematics #LinearRegression #Kaggle #DataScience #FinalPresentation #StatisticalAnalysis
=============
陳映竹！ (Chen, Ying-Chu)，Ph.D. candidate, M.S., M.A., CSCS*D, ACSM-CEP
=============
陳映竹同時擁有臺師大國際漢學研究所文學碩士，以及臺北市立大學運動健康科學系理學碩士。目前為臺師大體育與運動科學博士候選人。並且擁有美國運動醫學會臨床運動生理師以及私人教練認證！希冀透過平台持續推廣運動與健康科學相關知識，以及肌力與體能相關訓練理論及實務，並提供個人跨領域與教練自我成長歷程，讓有志從事教練產業以及斜槓工作者有所依循。
如果您覺得本次節目對您有幫助的話，就幫我按個讚喔！還有轉貼分享！
★★聯絡資訊 ★★
🟦 微小起心動念，將有莫大改變！小額捐款贊助運科博士候選人！
http://tinyurl.com/3jan2d46
🟦需統編者，請改點下面連結，可輸入統一編號！
https://vocus.cc/salon/stevelovesworkout/plans/content
🟨 自媒體經營：以「我是運科竹」為關鍵字搜尋！
▌Youtube: https://www.youtube.com/@coachsteveperformance
▌Instagram : https://www.instagram.com/coachsteveperformance/
▌Facebook : https://www.facebook.com/SteveLovesWorkout/
▌SubStack :https://coachsteveperformance.substack.com/
Podcast 頻道 : Sound on, Apple Podcast, 以及 Google Podcast，均為「我是運科竹」。
🟪 文章創作：
▌Vocus 方格子文章創作：https://vocus.cc/user/61d96b17fd89780001e2c067
▌Medium 文章創作：https://medium.com/@coachstevetw
🟦 個人網站：
▌Website: https://coachsteveperformance.wordpress.com/about/
▌CV website: https://lunacysaint.github.io
🟨 需要公益演講、商業合作，或者贊助運科博士候選人事宜，請電郵聯繫！
▌E-mail : [email protected]
