Strategic network portfolio

P6.2 selects at most one strategic option for each segment and period under period budgets, method capacities, a service-level target, a risk target, and a spatial-continuity preference.

Inputs

PortfolioSegment records a segment, its span and sequence, length, and no-treatment service, risk, and uncertainty. PortfolioPeriod records a year, budget, and capacity by method. PortfolioOption records a segment/period option, action, method, cost, capacity units, service, risk, and uncertainty.

Every option references a known segment and period. Option ids are unique. A no-treatment option must have zero cost. The committed example is built in the tests; there is no separate portfolio fixture in P6.2.

Selection

build_strategic_portfolio runs three deterministic passes:

  1. Service repair. While aggregate service is below service_target, pick the feasible option with the largest service gain per cost. Skip zero or negative gains.

  2. Objective improvement. While a feasible option improves the weighted objective, add it.

  3. Local swap. Replace a selected option with a better feasible option for the same segment and period.

Budget and method capacity are checked before every selection. The plan reports every hard violation.

Objective

service_weight * service
- risk_weight * risk
- uncertainty_weight * uncertainty
- cost_weight * total_cost
- continuity_weight * (1 - continuity_score)

service, risk, and uncertainty are length- and period-weighted. continuity_score is the share of adjacent segment pairs in a span that both have a selected option in the same period. The scalar objective never replaces the separate service, risk, uncertainty, cost, and feasibility fields.

Constraints and reports

  • Budget and capacity violations are hard PORTFOLIO-BUDGET and PORTFOLIO-CAPACITY errors.

  • A service miss is the hard PORTFOLIO-SERVICE error.

  • A risk target miss is the soft PORTFOLIO-RISK warning.

  • period_summary records cost, budget, capacity used, service, risk, uncertainty, and selection count per period.

  • selections records the selected option and its outcome.

  • The baseline service, risk, and uncertainty are reported for comparison.

Limits

  • The greedy construction and local swap are heuristic, not a global optimum.

  • The options and outcomes are supplied evidence; P6.7 wires them to the synthetic corridor.

  • seed is a scenario identifier; the current construction is deterministic without randomness.

  • A risk target is soft; a decision-maker must accept or reject the reported miss.