NIA Simulator — Nature-Inspired Algorithms

fitness function

PYTHON Telco churn — pilih offer terbaik per pelanggan berisiko

def fitness(solution, customers,
          budget):

          # solution = [offer_id per pelanggan], misal [2, 0, 1, 3, ...]

          total_cost = sum(OFFER_COST[solution[i]] for i in range(len(customers)))

          if total_cost > budget:

          return -999_999 # constraint
            violation

          retained = sum(

          customers[i]['arpu'] * ACCEPT_RATE[solution[i]][customers[i]['risk']]

          for i in range(len(customers))

          )

          return retained # maximise retained ARPU

          # OFFERS: 0=tanpa offer, 1=diskon 10%, 2=gratis 1 bulan, 3=priority support

          # GA akan mencari kombinasi terbaik yang memaksimalkan retained revenue

parameter simulasi

Pelanggan berisiko30 orang

Budget retensiRp 30jt

Ukuran populasi60

Generasi80

hasil optimasi

Revenue retained (GA)
—
—

Baseline (random)

—

assignment acak

Improvement vs baseline

—

GA vs random

Konvergen di generasi

—

CONVERGENCE CURVE — fitness per generasi

OPTIMIZATION LOG

Tekan "Jalankan GA" untuk memulai optimasi...

distribusi offer (hasil terbaik)

OFFER ALLOCATION TABLE

Belum ada hasil. Jalankan GA terlebih dahulu.

fitness function

PYTHON Penempatan tower — minimasi biaya + latency, coverage ≥ 95%

def fitness(solution):

          towers = solution.reshape(-1, 2) #
            [(x1,y1), (x2,y2), ...]

          coverage = compute_coverage(towers, DEMAND_POINTS, radius=3.0)

          if coverage < 0.95:

            return -999_999 # wajib cover 95%
              demand

            avg_latency = mean_nearest_distance(DEMAND_POINTS, towers)

            deploy_cost = len(towers) * BIAYA_PER_TOWER # Rp
              500jt/tower

            return -(avg_latency * 0.6 + deploy_cost * 0.4) # minimise

            # PSO: setiap partikel = satu kandidat penempatan N tower

            # velocity update: v = w*v + c1*r1*(pBest-x) + c2*r2*(gBest-x)

parameter simulasi

Jumlah tower6 tower

Titik demand50 titik

Jumlah partikel25

Biaya/towerRp 500jt

hasil optimasi

Coverage demand
—
target ≥ 95%

Avg latency

—

jarak ke tower terdekat

Total capex

—

PETA PENEMPATAN TOWER — area 10×10 km

Titik demand tercovered

Tidak tercovered

Tower BTS

Coverage radius

OPTIMIZATION LOG

Tekan "Jalankan PSO" untuk memulai optimasi...

fitness function

PYTHON Jadwal agen — minimasi biaya + penalti SLA violation

def fitness(schedule, agents, shifts,
          sla_min):

          # schedule[i] = shift yang ditugaskan ke agen i

          cost = sum(agents[a]['rate'] * shifts[s]['hours']

          for a, s in enumerate(schedule))

          coverage = defaultdict(int)

          for a, s in enumerate(schedule):

          for h in shifts[s]['covers']:
          coverage[h] += 1

          penalty = sum(

          max(0, sla_min - coverage[h]) * 500_000 # Rp 500rb per slot kurang

          for h in ALL_HOURS

          )

          return -(cost + penalty) # minimise total

parameter simulasi

Jumlah agen14 agen

SLA min per slot2 agen/jam

Rate honor/jamRp 35rb

Generasi100

hasil optimasi

Biaya mingguan
—
—

SLA violations

—

slot jam di bawah target

Avg coverage

—

agen per slot jam

STAFFING HEATMAP — coverage agen per jam— SLA minimum

CONVERGENCE CURVE

OPTIMIZATION LOG

Tekan "Jalankan GA" untuk memulai optimasi...

distribusi shift (hasil terbaik)

Belum ada hasil.