Ranking score. Search algorithm logic
Machine Learning Ranking Model
Model predicts:
βProbability this provider will accept and complete the job fastest.β
Train a model using:
Distance data:
Response time
Rating
Price
Acceptance rate
Completion rate
Time of day
Road type
ranking engine tailored for recovery providers, distanceβbased dispatch
Ranking Formula (Weighted Score)
Your ranking score should combine:
Distance score
Response time score
Rating score
Price score
Verification score
Availability score
Final Score Formula
score
=
(
0.40
β
π
π
)
+
(
0.20
β
π
π
)
+
(
0.20
β
π
π
)
+
(
0.10
β
π
π
)
+
(
0.10
β
π
π£
)
Where:
π
π
= distance score
π
π
= response time score
π
π
= rating score
π
π
= price score
π
π£
= verification score
RealβWorld Dispatch Enhancements
A. Surge Pricing / Busy Area Boost
If many users are searching in the same area:
Increase weight of response time
Decrease weight of price
B. Provider Fatigue Protection
If a provider has accepted too many jobs recently:
Temporarily reduce their score
C. Fraud Prevention
If a provider rejects too many jobs:
Lower ranking
Trigger admin review
D. AIβBased ETA Prediction
Use:
Traffic
Time of day
Provider historical speed
To calculate a more accurate ETA.