Insights on AI strategy, modelling, and what it takes to put state-of-the-art models into production.分享 AI 策略、模型建構，以及將先進模型真正落地到生產環境所需的實務經驗。
A practical comparison of linear regression and XGBoost across 120 store–SKU series, testing whether promotions, pricing, marketing, competitors, and weather improve one-day-ahead forecasts.以 120 條門市－SKU 序列比較線性迴歸與 XGBoost，檢驗促銷、定價、行銷、競品與天氣資訊能否改善隔日需求預測。
18 min read閱讀約 26 分鐘
A practical comparison of a global LSTM with XGBoost across 120 store–SKU series, asking whether learned temporal representations justify the added complexity.以 120 條門市－SKU 序列比較全域 LSTM 與 XGBoost，評估學習式時間表徵能否帶來足以抵銷額外成本的預測改善。
18 min read閱讀約 28 分鐘
A practical and theoretical guide to forecasting iPhone demand with Holt–Winters and SARIMA—from backtesting and model selection to point forecasts.以 iPhone 隔日需求為例，實作 Holt–Winters 與 SARIMA，完整說明時間序列回測、模型選擇、殘差診斷與預測區間。
20 min read閱讀約 34 分鐘
An Apple Store demand forecasting problem, solved three ways—Statistics, Machine Learning, and LSTM—to explore which model truly earns its complexity.以 Apple Store 的需求規劃為例，比較統計模型、監督式機器學習與 LSTM，說明不同方法在資料規模與商業情境下的適用條件。
14 min read閱讀約 18 分鐘