RESPONSE MODEL · CINEMA PROGRAMMING
Optimize the screening timetable
Import a reference schedule, set the operating rules, and compare the best feasible timetable found by the response optimizer.
Import the timetable to improve
Load a complete planning package with films, screens, response assumptions and a current timetable. You can also replace the timetable with a CSV, using the currently loaded film and screen definitions.
Try a complete classroom planning case
Six films, six sites, 18 screens and one operating day. The sample represents a feasible current timetable that can be improved.
JSON supplies the complete dataset. CSV requires showing_id, site_id, screen_id, film_id and start_time; it uses the loaded package, or the classroom definitions if none is loaded. Maximum file size: 2 MiB.
What does the planning file contain?
Film runtimes and room eligibility; site screen capacities and film tastes; declared film/time response factors; operating and coverage rules; and the current screening start times. These planning inputs are supplied separately from purchase records. This demo evaluates an explicit response rule; it does not fit that rule from uploaded purchases.
The supported shape is six film IDs A–F and six site IDs A–F, with three screens at each site. A complete reference uses every screen and must pass its operating, film-cap, coverage and room-fit rules.
Review the inputs and set the operating rules
The uploaded timetable is the reference for this run. Confirm the response assumptions and constraints before optimizing.
Optimization objective
Each screening's prediction is capped by its seats. The objective sums those predictions and rounds once.
A changed rule must still admit the uploaded reference timetable. If it does not, correct the timetable or the rule before continuing. The optimizer then starts from that exact reference; it does not run a hidden presolve.
Film definitions and supply caps
| Film | Runtime | Base response | Day / prime | Chain cap | Screen eligibility |
|---|
Sites, screens and audience tastes
| Site | Screen capacities | Film tastes / A–F |
|---|
Uploaded reference timetable
| Showing ID | Site | Screen | Film | Start | Runtime |
|---|
Response model assumptions
Response combines the film base, site taste, time weight and day/prime factors. Nearby same-film starts within 90 minutes multiply it by 0.72 each; other-film starts within 20 minutes multiply it by 0.94 each. Additional same-film shows at the site multiply it by 0.93 each. Large-screen eligibility and bonuses come from the planning package.
Search for a better feasible timetable
Optimize film, screen and time assignments from the confirmed reference. Every accepted candidate uses the same response model and operating rules.
Local ILS checks operating times, cleaning, screen eligibility, film caps and site coverage before response evaluation. The retained plan is the best feasible timetable found within the selected search budget.
This run evaluates declared synthetic response assumptions. It does not fit a model from purchases or run Bayesian optimization.
Compare and inspect the timetable
Current best ·
Inspect what changed
Confirmed search start
Best feasible timetable found
Explore a different programming decision
Exchange two films or add a gap, inspect the response change, then apply the feasible experiment to prepare another search. Drafts start from the uploaded reference.
Explore a film exchange
Choose two screenings and swap their films. Each film keeps its total number of shows. Different runtimes move later screenings on the affected screens; the response model evaluates the complete updated timetable.
Start with 0. Add a gap to test the value of a later time slot. Every draft begins from the original timetable.
Original timetable · affected screens
Your draft · affected screens
| Screening slot | Film · original → draft | Start / end · original → draft | Expected attendance | Change |
|---|
The table includes every affected prediction, including screenings whose own film and time stay the same but whose nearby competition changes. Whole-plan scores round once; row predictions use one decimal.
Why does the response change?
Compare the model inputs for the two exchanged slots. Same-film starts within 90 minutes multiply response by 0.72 each; other-film starts within 20 minutes multiply it by 0.94 each. Additional same-film shows at the site multiply it by 0.93 each. Film C may receive a large-screen factor. The final prediction is capped by the screen’s seats. These are declared synthetic assumptions.
| Exchanged slot | Film base × site taste | Time weight × period | Nearby same / other films | Daily same-film shows | Large-screen factor / seat cap |
|---|
Auxiliary checks · operating-hours boundary and rejection
Test a closing-time constraint
Predictions and gains are conditional on the declared response rule. They are synthetic expected attendance across screenings, not observed admissions or unique viewers. A search result is not a proven global optimum. Read the English demo guide.