Logo
0

找预分单运力——>判断预分单or升舱

当前轮询逻辑：不放回（不累加）可能有问题

供需判断——>找预分单运力——>判断预分单or升舱or不预分单（举手后）

  1. 供需判断：当前格子内供需情况较差，触发预分单/升舱
  2. 判断对应商家升舱/预分单 卡点
  3. 判断之后，让对应商家其寻找司机运力 a. 升舱：找到运力后直接完单 ❗没找到运力怎么办 b. 不升舱：让其经过预分单接受度模型，接受度高于阈值的话进行预分单，否则什么都不做

总目标：增加完单

预算：0.1pp cost

完单：0.4pp

参数与变量定义

Pi(A∣c)P_{i}(A|c)

PiupP_{i}^{up}

PibaseP_{i}^{base}

priceiupprice_{i}^{up}

priceibaseprice_{i}^{base}

GMViupGMV_{i}^{up}

GMVibaseGMV_{i}^{base}

ziz_{i}

θ\theta

目标函数

maxf(Z)=∑i=1Nzi(Piup−[Pi(A∣c)Piup+(1−Pi(A∣c))Pibase)])\begin{gather} maxf(\mathbf{Z})=\sum_{i=1}^{N}{{z_{i} (P_{i}^{up}-[P_{i}(A|c) P_{i}^{up}+(1-P_{i}(A|c)) P_{i}^{base})])}} \end{gather}

约束条件

s.t∑i=1Nzi(priceiup−priceibase)Piup≤BB=(∑i=1Nzi[(1−Pi(A∣c))(GMViup−GMVibase)]+(1−Pi(A∣c))GMVibase+Pi(A∣c)GMViup)⋅θ\begin{gather} s.t \quad \quad \sum_{i=1}^{N}{z_{i}(price_{i}^{up}-price_{i}^{base})P_{i}^{up}} \leq B \\ B=(\sum_{i=1}^{N}{z_{i}[(1-P_{i}(A|c))(GMV_{i}^{up}-GMV_{i}^{base})]+ (1-P_{i}(A|c))GMV_{i}^{base}+P_{i}(A|c)GMV_{i}^{up})} \cdot \theta \end{gather}

优化决策

minL(Z,λ)=−∑i=1Nzi(Piup−[Pi(A∣c)Piup+(1−Pi(A∣c))Pibase])+λ⋅(∑i=1Nzi(priceiup−priceibase)Piup−B)\begin{gather} minL(\mathbf{Z},\lambda)=-\sum_{i=1}^{N}{{z_{i} (P_{i}^{up}-[P_{i}(A|c) P_{i}^{up}+(1-P_{i}(A|c)) P_{i}^{base}])}}\\+\lambda \cdot(\sum_{i=1}^{N}{z_{i}(price_{i}^{up}-price_{i}^{base})P_{i}^{up}}-B) \end{gather} min⁡Zmax⁡λL(Z,λ)\begin{gather} \min_{\mathbf{Z}}{\max_{\lambda}}L(\mathbf{Z},\lambda) \end{gather} max⁡λmin⁡ZL(Z,λ)=−λB+∑i=1Nzi[λ(priceiup−priceibase)Piup−(1−Pi(A∣c)(Piup−Pibase)]⏟连续背包问题中的valueweight=−λ(∑i=1Nzi[(1−Pi(A∣c))(GMViup−GMVibase)]+(1−Pi(A∣c))GMVibase+Pi(A∣c)GMViup)⋅θ+∑i=1Nzi[λ(priceiup−priceibase)Piup−(1−Pi(A∣c)(Piup−Pibase)]⏟连续背包问题中的valueweightmax⁡λmin⁡ZL(Z,λ)=∑i=1Nzi[[1−Pi(A∣c)][−λθ(GMViup−GMVibase)−(Piup−Pibase)]+λ(priceiup−priceibase)Piup)]⏟连续背包问题中的valueweight−∑i=1Nλθ[(1−Pi(A∣c))GMVibase+Pi(A∣c)GMViup]\begin{gather} \max_{\lambda}{\min_{\mathbf{Z}}}L(\mathbf{Z},\lambda)=-\lambda B+\sum_{i=1}^{N}{z_{i} \underbrace{[\lambda (price_{i}^{up}-price_{i}^{base})P_{i}^{up}-(1-P_{i}(A|c)(P_{i}^{up}-P_{i}^{base})]}_{连续背包问题中的\frac{value}{weight}}} \\ =-\lambda (\sum_{i=1}^{N}{z_{i}[(1-P_{i}(A|c))(GMV_{i}^{up}-GMV_{i}^{base})]+ (1-P_{i}(A|c))GMV_{i}^{base}+P_{i}(A|c)GMV_{i}^{up})} \cdot \theta \\ +\sum_{i=1}^{N}{z_{i} \underbrace{[\lambda (price_{i}^{up}-price_{i}^{base})P_{i}^{up}-(1-P_{i}(A|c)(P_{i}^{up}-P_{i}^{base})]}_{连续背包问题中的\frac{value}{weight}}} \\ \max_{\lambda}{\min_{\mathbf{Z}}}L(\mathbf{Z},\lambda)=\sum_{i=1}^{N}{z_{i} \underbrace{[[1-P_{i}(A|c)][-\lambda \theta (GMV_{i}^{up}-GMV_{i}^{base})-(P_{i}^{up}-P_{i}^{base})]+\lambda (price_i{^{up}-price_{i}^{base})P_{i}^{up}})]}_{连续背包问题中的\frac{value}{weight}}} \\ -\sum_{i=1}^{N}{\lambda \theta [(1-P_{i}(A|c))GMV_{i}^{base}+P_{i}(A|c)GMV_{i}^{up}]} \end{gather}

贪心计算指标：

[[1−Pi(A∣c)][λθ(GMViup−GMVibase)+(Piup−Pibase)]−λ(priceiup−priceibase)Piup]\begin{gather} [[1-P_{i}(A|c)][\lambda \theta (GMV_{i}^{up}-GMV_{i}^{base})+(P_{i}^{up}-P_{i}^{base})]-\lambda (price_{i}^{up}-price_{i}^{base})P_{i}^{up}] \end{gather}

贪心计算指标（原）：

(1−Pi(A∣c))(Piup−Pibase)−λ(priceiup−priceibase)Piup(1-P_{i}(A|c))(P_{i}^{up}-P_{i}^{base}) -\lambda (price_{i}^{up}-price_{i}^{base})P_{i}^{up}

根据该升舱规划模型判断是否升舱，若要升舱的话再去让商家找司机（不看距离），如果不升舱再过预分单接受度模型

© 2025 All rights reservedBuilt with DataHub Cloud

Built with LogoDataHub Cloud