课程大纲设计指南

课程名称：数据科学导论（16周）/ Course Title: Introduction to Data Science (16 Weeks)

1. 课程概览 / Course Overview

• 课程定位：面向初学者的基础课程，覆盖数据科学全流程：获取与清洗、探索与可视化、统计推断、机器学习入门、结果沟通与复现。/ Scope: Foundational course covering the end-to-end data science lifecycle: data acquisition/cleaning, EDA/visualization, statistical inference, introductory machine learning, communication and reproducibility.
• 教学形式：讲授 + 实操实验课 + 项目。/ Modality: Lectures + hands-on labs + project.
• 接触时数：每周3小时（2小时讲授 + 1小时实验/讨论），16周。/ Contact hours: 3 per week (2h lecture + 1h lab/discussion), 16 weeks.
• 适用对象与先修：具备基础编程（任一语言）与高中层次代数/函数知识；不要求先修统计学。/ Audience & prerequisites: Basic programming (any language) and high-school algebra/functions; no prior statistics required.
• 教学语言与提交：代码与报告可用中文或英文；建议英文代码注释以便通用复现。/ Language & submissions: Chinese or English; recommend English code comments for portability.

2. 课程学习成果（CLOs）/ Course Learning Outcomes 完成课程后，学生能够 / Upon completion, students will be able to:

• 描述数据科学工作流与常见角色、任务与伦理边界 / Describe DS workflow, roles, tasks, and ethical boundaries.
• 使用Python、NumPy、pandas进行数据摄取、清洗、变换与合并 / Use Python, NumPy, pandas for ingestion, cleaning, transformation, merging.
• 进行探索性数据分析并构建有效可视化 / Perform EDA and create effective visualizations.
• 以SQL查询关系型数据并进行分组聚合与连接 / Query relational data with SQL including grouping and joins.
• 应用概率与统计推断（估计、区间、假设检验、A/B测试） / Apply probability and inference (estimation, CIs, hypothesis tests, A/B testing).
• 训练与评估基础监督学习模型（回归与分类），并执行交叉验证与调参 / Train and evaluate baseline supervised models (regression/classification) with CV and tuning.
• 应用聚类与PCA等无监督方法进行模式发现与降维 / Use clustering and PCA for pattern discovery and dimensionality reduction.
• 构建可复现的分析工作流（Git、环境管理、流水线）并清晰沟通结果 / Build reproducible workflows (Git, environments, pipelines) and communicate findings.

3. 教学材料与工具 / Materials and Tools

• 编程环境 / Programming: Python 3.10+; JupyterLab; Conda或venv。/ JupyterLab; Conda or venv.
• 主要库 / Core libraries: NumPy, pandas, matplotlib, seaborn, scikit-learn; statsmodels（可选）。/ Optional: statsmodels.
• 数据库 / Database: SQLite（必备）; PostgreSQL（可选）。/ SQLite required; PostgreSQL optional.
• 版本控制 / Version control: Git/GitHub或等效平台。/ Git/GitHub or equivalent.
• 文档与报告 / Documentation: Jupyter Notebook/Quarto；Streamlit（可选）用于展示。/ Streamlit optional for demos.
• 推荐教材与参考 / Recommended references:
  • Wes McKinney, Python for Data Analysis, 3rd ed., 2022.
  • Gareth James et al., An Introduction to Statistical Learning with Applications in Python, 2023.
  • Aurélien Géron, Hands-On Machine Learning with Scikit-Learn, Keras & TensorFlow, 3rd ed., 2022.
  • Peter Bruce, Andrew Bruce, Peter Gedeck, Practical Statistics for Data Scientists, 3rd ed., 2023.
  • 官方文档：pandas、scikit-learn、seaborn、SQLite。/ Official docs for pandas, scikit-learn, seaborn, SQLite.

4. 评估组成与权重 / Assessment and Weights

• 课堂与实验参与 Participation/Labs: 10%
• 小测（4次） Quizzes (4): 10%
• 编程作业（6次） Programming Assignments (6): 20%
• 期中考试 Midterm Exam: 20%
• 课程项目 Capstone Project（含提案、里程碑、报告与展示）: 30%
• 同行评审 Peer Review: 10% 说明：强调学术诚信与可追溯性；代码需可运行、结果可复现。/ Note: Emphasis on academic integrity and reproducibility; code must run and results be reproducible.

5. 每周进度与任务 / Weekly Schedule and Tasks

第1周：课程导入与环境搭建 / Week 1: Orientation and Environment Setup

• 主题/Topic：数据科学生态、工作流、工具栈；Jupyter、Git、环境。/ DS landscape, workflow, stack; Jupyter, Git, environments.
• 目标/Objectives：完成环境安装；能运行Notebook；理解课程期望与项目要求。/ Install environment; run notebooks; understand course and project.
• 实验/Lab：环境检查、Git基础、数据读写。/ Env check, Git basics, I/O.
• 作业/HW：工具链再现报告（含环境说明与样例分析）。/ Reproducible setup report with sample analysis.

第2周：Python与NumPy基础 / Week 2: Python and NumPy Essentials

• 主题：数据结构、控制流、函数、向量化与广播。/ Data structures, control flow, functions, vectorization.
• 目标：编写可读、可测试函数；使用NumPy进行高效计算。/ Write testable functions; use NumPy efficiently.
• 实验：数值计算与基准比较。/ Numeric computing and benchmarking.
• 作业：实现数据清洗实用函数库。/ Implement utility functions for cleaning.

第3周：pandas数据整理 / Week 3: Data Wrangling with pandas

• 主题：索引/选择、缺失值、连接/合并、重塑（melt/pivot）。/ Indexing, missing values, joins/merge, reshape.
• 目标：构建整洁数据；记录数据字典。/ Produce tidy data; document a data dictionary.
• 实验：清洗“脏数据”案例。/ Clean a messy dataset.
• 作业：完成数据清洗与整合报告。/ Deliver a cleaning/integration report.

第4周：探索性数据分析与可视化 / Week 4: EDA and Visualization

• 主题：分布、关系、异常点、图形原则。/ Distributions, relationships, outliers, viz principles.
• 目标：形成可复现EDA；选择恰当图表。/ Reproducible EDA; appropriate chart selection.
• 实验：用seaborn制作EDA图谱。/ EDA gallery with seaborn.
• 作业：EDA Notebook与洞见摘要。/ EDA notebook and insight brief.

第5周：数据伦理与复现 / Week 5: Data Ethics and Reproducibility

• 主题：隐私、偏见、公平、同意；随机种子、日志、依赖锁定。/ Privacy, bias, fairness, consent; seeds, logging, dependency locking.
• 目标：完成伦理清单；建立可复现项目模板。/ Ethics checklist; reproducible project template.
• 实验：偏见风险初筛；requirements.txt生成。/ Bias pre-mortem; requirements.txt.
• 项目里程碑/Milestone：选题与数据源确认 + 伦理清单草案。/ Topic and data source + ethics checklist draft.

第6周：概率基础与模拟 / Week 6: Probability Foundations and Simulation

• 主题：随机变量、期望/方差、常见分布、LLN/CLT（概念）、蒙特卡罗。/ RVs, moments, common distributions, LLN/CLT (conceptual), Monte Carlo.
• 目标：用模拟验证概率结论；选择合适分布建模。/ Validate via simulation; choose distributions.
• 实验：置信区间覆盖率模拟。/ CI coverage simulation.
• 作业：概率建模与模拟报告。/ Probability modeling report.
• 项目里程碑：项目提案1页（问题、数据、指标、风险）。/ 1-page proposal (problem, data, metrics, risks).

第7周：统计推断与实验设计 / Week 7: Statistical Inference and Experimentation

• 主题：抽样、点估计与区间、t检验/卡方、p值与效应量、检验力、A/B测试。/ Sampling, CIs, t/chi-square tests, p-values, effect size, power, A/B.
• 目标：构建与解释假设检验；避免常见谬误。/ Build and interpret tests; avoid pitfalls.
• 实验：A/B功效分析与结果解读。/ A/B power analysis and interpretation.
• 作业：实验设计备忘录。/ Experiment design memo.

第8周：SQL与期中 / Week 8: SQL + Midterm

• 主题：SELECT、WHERE、GROUP BY、HAVING、JOIN、子查询；数据摄取与API概览。/ SELECT, WHERE, GROUP BY, HAVING, JOINs, subqueries; ingestion and APIs overview.
• 目标：从关系数据库提取特征数据集。/ Extract feature-ready datasets from RDBMS.
• 实验：SQLite业务查询实战。/ SQLite query lab.
• 评估/Assessment：期中考试（覆盖第1–7周）。/ Midterm exam (Weeks 1–7).

第9周：监督学习I—回归 / Week 9: Supervised Learning I—Regression

• 主题：一元/多元线性回归、假设与诊断、正则化（Ridge/Lasso）入门。/ Linear/multiple regression, assumptions/diagnostics, intro to Ridge/Lasso.
• 目标：拟合与解读回归；处理多重共线性与过拟合。/ Fit/interpret; handle multicollinearity and overfitting.
• 实验：房价或类似数据回归。/ Housing-like regression lab.
• 作业：回归建模与残差分析。/ Regression + residual analysis.
• 项目里程碑：EDA初稿提交。/ EDA draft due.

第10周：监督学习II—分类 / Week 10: Supervised Learning II—Classification

• 主题：逻辑回归、kNN；混淆矩阵、精确率/召回率、ROC-AUC；类别不平衡。/ Logistic regression, kNN; confusion matrix, precision/recall, ROC-AUC; class imbalance.
• 目标：选择恰当评估指标；应对不平衡。/ Choose metrics; address imbalance.
• 实验：欺诈/流失分类任务。/ Fraud/churn classification.
• 作业：分类器比较与报告。/ Classifier comparison report.

第11周：树模型与集成 / Week 11: Trees and Ensembles

• 主题：决策树、随机森林、梯度提升；变量重要性与可解释性概念。/ Decision trees, Random Forest, Gradient Boosting; feature importance and interpretability concepts.
• 目标：训练稳健基线；理解偏差-方差权衡。/ Train robust baselines; bias-variance trade-off.
• 实验：随机森林调参。/ Random Forest tuning.

第12周：模型评估与流水线 / Week 12: Model Evaluation and Pipelines

• 主题：训练/验证/测试分割、交叉验证、数据泄漏、预处理（缩放/编码）、管道、超参搜索。/ Train/val/test splits, CV, leakage, preprocessing (scaling/encoding), pipelines, hyperparameter search.
• 目标：构建端到端可复现ML流水线。/ Build end-to-end reproducible ML pipelines.
• 实验：scikit-learn Pipeline + GridSearchCV。/ Pipelines + GridSearchCV.
• 项目里程碑：建模计划与原型。/ Modeling plan + prototype.

第13周：无监督学习 / Week 13: Unsupervised Learning

• 主题：k-means、层次聚类、轮廓系数；PCA；高维可视化注意事项（t-SNE/UMAP概念）。/ k-means, hierarchical clustering, silhouette; PCA; high-dim viz caveats (t-SNE/UMAP concept).
• 目标：执行聚类并评估可分性；使用PCA降维与解释方差。/ Run clustering; use PCA and explain variance.
• 实验：客户分群或文本TF-IDF聚类（可选）。/ Customer segmentation or TF-IDF clustering (optional).

第14周：时间序列与特征工程 / Week 14: Time Series and Feature Engineering

• 主题：日期时间处理、重采样、滚动窗口、平稳性直觉、简单基线预测；特征工程与编码策略。/ Datetime handling, resampling, rolling windows, stationarity intuition, baseline forecasting; feature engineering and encoding strategies.
• 目标：构建时间序列基线与特征；避免数据泄漏。/ Build TS baselines and features; avoid leakage.
• 实验：销量或传感器数据基线预测。/ Sales/sensor baseline forecast.
• 作业：特征工程备忘录。/ Feature engineering memo.

第15周：沟通、可视化叙事与复现交付 / Week 15: Communication, Storytelling, and Reproducible Delivery

• 主题：受众导向的信息架构；图表选择；不确定性表达；报告与仪表板；环境与数据卡。/ Audience-focused narratives; chart choices; uncertainty communication; reports/dashboards; environment and data cards.
• 目标：完成可复现报告与演示材料。/ Finalize reproducible report and presentation.
• 实验：用Quarto/Jupyter或Streamlit打包成果。/ Package deliverables with Quarto/Jupyter or Streamlit.
• 项目里程碑：项目草稿与同伴演练。/ Draft and peer rehearsal.

第16周：项目展示与课程总结 / Week 16: Final Presentations and Wrap-up

• 评估：项目口头展示、代码审阅、最终报告提交。/ Assessment: Oral presentations, code review, final report submission.
• 总结：课程回顾、扩展学习路径建议。/ Wrap-up and pathways for further study.

6. 课程项目要求 / Capstone Project Requirements

• 交付物/Deliverables：1页提案、伦理清单、EDA报告、建模原型、最终报告（≤10页）与演示（8–10分钟）、可运行代码与环境文件。/ 1-page proposal, ethics checklist, EDA report, modeling prototype, final report (≤10 pages) and 8–10 min presentation, runnable code with environment file.
• 评分维度/Rubric dimensions：问题定义与价值（20%）、数据质量与处理（15%）、方法与严谨性（30%）、结果与不确定性沟通（20%）、复现与代码质量（15%）。/ Problem/value (20%), data quality/wrangling (15%), methods/rigor (30%), communication of results/uncertainty (20%), reproducibility/code (15%).
• 数据来源/Data sources：公开数据集（如UCI、政府开放数据等）或教师批准的自选数据；遵循使用条款与隐私规范。/ Public datasets (e.g., UCI, open government) or approved data; comply with ToS and privacy.

7. 教学与学习策略 / Instructional and Learning Strategies

• 翻转+实作：课前微视频与阅读；课堂聚焦练习与代码演示。/ Flipped micro-lectures + in-class coding.
• 渐进复杂度：从清洗与EDA过渡到建模与评估。/ Progressive complexity from wrangling/EDA to modeling/evaluation.
• 形成性评估：每周短测与即时反馈。/ Frequent formative quizzes and feedback.
• 同伴互评：提高沟通质量与代码可读性。/ Peer review to improve communication and code readability.

8. 学术诚信与可访问性 / Academic Integrity and Accessibility

• 学术规范：独立完成个人作业；允许讨论但需署名引用；禁止抄袭与未经授权的模型/代码共享。/ Individual assignments; discussion allowed with attribution; no plagiarism or unauthorized code/model sharing.
• 可访问性：提供可机读资料、字幕与大纲；为必要的学习便利提供支持。/ Accessible materials (machine-readable, captions); reasonable accommodations available.

9. 建议每周阅读（与主题对应）/ Suggested Weekly Readings (aligned to topics)

• W1–W2: Python for Data Analysis (PfDA) Ch. 1–4.
• W3: PfDA Ch. 5–8（数据整理与重塑）。/ Wrangling and reshape.
• W4: seaborn/matplotlib 官方指南；数据可视化最佳实践文章。/ Official guides; viz best practices.
• W5: 负责任AI与数据伦理入门文献；Git与可复现研究实践。/ Responsible AI/data ethics; reproducible research practices.
• W6–W7: Practical Statistics for Data Scientists（概率/推断章节）。/ Probability and inference chapters.
• W8: SQL官方教程或SQL入门教材。/ SQL tutorials.
• W9–W12: ISL with Python（回归/分类/模型评估章节）；Hands-On ML（模型评估与管道）。/ ISLP regression/classification/evaluation; Hands-On ML pipelines.
• W13–W14: ISLP/Hands-On ML（无监督与降维；时间序列基础参考资料）。/ Unsupervised and dimensionality reduction; TS basics references.
• W15: 数据叙事与技术写作指南；Streamlit/Quarto文档。/ Data storytelling and technical writing; Streamlit/Quarto docs.

10. 评估基准与通过标准 / Performance Benchmarks

• 通过标准：总评≥60%且项目≥60%；期中考试≥50%。/ Pass if overall ≥60% and project ≥60%; midterm ≥50%.
• 代码质量：通过自动化测试/样例数据运行；随机种子固定；读我文档与依赖文件完整。/ Code runs on sample tests; random seeds fixed; README and dependency files complete.

备注 / Notes

• 所列教材与工具为业界主流与稳定版本，适合入门课程；若版本更新不影响核心API与概念，课程材料将同步微调。/ Listed texts and tools are mainstream and stable for introductory courses; minor updates will be accommodated without changing core APIs and concepts.
• 本大纲为基础框架，授课进度可根据学习诊断与班级背景进行微调。/ This outline is a baseline; pacing may be adjusted based on diagnostics and cohort background.