Demand Forecasting
Mathematical Formula & Modeling
The transformation of a datetime:
1. Mathematical Representation
Let T be the target timestamp and torigin be the reference point. The value in decimal days D is defined as:
Where:
- fsec: A function that extracts the total accumulated seconds from a time duration.
- 86,400: The constant for total seconds in a day (24 hours × 3,600 seconds).
- D: A continuous scalar in ℝ (Real Numbers).
2. Logical Modeling Process
In data modeling, this transformation follows a Linear Temporal Scaling logic:
| Stage | Operation | Result Type |
|---|---|---|
| Translation | (T - torigin) | Timedelta Object |
| Quantification | .total_seconds() | Integer/Float (Seconds) |
| Normalization | / 86,400 | Normalized Float (Days) |
- Công thức tổng quát:
y =f(x) ︸ Quy luật (Signal)+ε ︸ Sai số (Noise)
- Constant:
K(x, x') = σ₀²
- Thời gian tuyến tính theo ngày:
Exponential function: exp(x) = e ^x
Trong đó:
- e là số Euler, một hằng số toán học xấp xỉ bằng
2,71828
- x là số mũ bạn đưa vào.
Mô tả:
- Hôm nay và hôm qua (hoặc cách đây n ngày) có liên quan gì nhau không. Ví dụ: Nếu ngày 01/02 kho rất đông khách, thì ngày 02/02 khả năng cao vẫn sẽ đông vì các chương trình khuyến mãi thường kéo dài vài ngày.
f = | x - x’|
Ý nghĩa: Giúp mô hình bắt được xu hướng ngắn hạn (ví dụ: một đợt cao điểm mua sắm kéo dài 1 tuần).
2. Ngày trong tuần" (Tính chu kỳ):
Decomposition of Multi-Component Gaussian Process Regression for Logistics Demand Forecasting
-
Overview:
The charts illustrate the daily pickup and delivery weights across three key regions: Ho Chi Minh City (HCMC), Hanoi, and Binh Thuan. A shared characteristic across all regions is a significant, sharp decline in volume during late January and early February 2025, which aligns with the Lunar New Year (Tet) holiday period.
-
Ho Chi Minh City & Hanoi: High-Volume Hubs:
-
In both cities, pickup weight (Khối lượng lấy) consistently exceeds delivery weight (Khối lượng giao). This suggests these cities act as major distribution points or manufacturing origins.
-
Both regions show high daily volatility, with HCMC peaking near 1.5 x 10^9 kg in September and Hanoi reaching its highest peak of approximately $1.6 x 10^9 kg in December
-
There is a visible upward trend in activity toward the end of the year (Q4), likely driven by shopping festivals and holiday demand.
Figure 1. Comparison of delivery and pickup weights across Ho Chi Minh City, Hanoi.
-
Binh Thuan: The Outlier Profile:
-
The volume in Binh Thuan is significantly lower, measured in tens of millions (10^7 kg) rather than billions.
-
Unlike HCMC and Hanoi, Binh Thuan's delivery weight (Khối lượng giao) far exceeds its pickup weight (Khối lượng lấy) throughout the year. This identifies Binh Thuan primarily as a "destination region" rather than a "source region."
Figure 2. Regional weight breakdown for delivery and pickup services of Binh Thuan Province
-
Logistics Demand Analysis Factors
Table 1.
-
Methodology
Figure
2. Composite Kernel Architecture
References:
Gaussian Processes for Machine Learning: Contents. (n.d.). https://gaussianprocess.org/gpml/chapters/