Online Learning & Drift Detection
প্রতি ৩ ঘণ্টায় স্বয়ংক্রিয় model retrain · উন্নত হলেই deploy
Retraining Strategy
Frequency
৩ ঘণ্টা
Auto scheduled
Drift Threshold
১০%
CPU বা RAM mean drift
Deploy Policy
উন্নত হলে
Only if RMSE improves
মোট Retrain
50
এখন পর্যন্ত
🚀 সফল Deploy
20
মোট 50 retrain এর মধ্যে
📉 সর্বশেষ CPU RMSE
7.43%
আগে ছিল: 7.25%
⚡ সর্বশেষ CPU Drift
290.6%
Threshold: 10% · V1003
সর্বশেষ Retrain এর প্রভাব — Real-time Database থেকে
🧠 CPU Forecast — V1003 · 2026-10-12 20:19
আগে
7.25%
RMSE
→
পরে
7.43%
RMSE
2% ↑
খারাপ
Drift: 290.6%
· R²: 0.774 → 0.763
· Samples: 45,125
💾 RAM Forecast — V1003 · 2026-10-12 20:19
আগে
1.02%
RMSE
→
পরে
1.02%
RMSE
0% ↓
উন্নত
RAM Drift: 36.9%
· Deploy: ✅ হয়েছে
RMSE উন্নতির গ্রাফ
CPU RMSE (%)
RAM RMSE (%)
Retrain ইতিহাস (সর্বশেষ 50টি)
| সময় | VPS | CPU আগে | CPU পরে | RAM আগে | RAM পরে | CPU Drift | Samples | Status |
|---|---|---|---|---|---|---|---|---|
| 2026-10-12 20:19 | V1003 | 7.2475% | 7.4264% ↑ | 1.0173% | 1.0169% ↓ | 290.6% | 45,125 | ✅ Deploy |
| 2026-10-12 17:18 | V1003 | 7.1658% | 7.2748% ↑ | 1.0177% | 1.0201% ↑ | 281.5% | 45,089 | ⏭️ Skip |
| 2026-10-12 14:18 | V1003 | 7.0954% | 7.2485% ↑ | 1.0178% | 1.0201% ↑ | 294.6% | 45,053 | ⏭️ Skip |
| 2026-10-12 11:17 | V1003 | 6.9980% | 7.2063% ↑ | 1.0208% | 1.0242% ↑ | 276.8% | 45,017 | ⏭️ Skip |
| 2026-10-12 08:17 | V1003 | 6.9079% | 6.9655% ↑ | 1.0206% | 1.0234% ↑ | 254.9% | 44,981 | ⏭️ Skip |
| 2026-10-12 05:16 | V1003 | 6.8495% | 6.9248% ↑ | 1.0198% | 1.0236% ↑ | 272.9% | 44,945 | ⏭️ Skip |
| 2026-10-12 02:16 | V1003 | 6.7872% | 6.9113% ↑ | 1.0201% | 1.0225% ↑ | 235.1% | 44,909 | ⏭️ Skip |
| 2026-10-11 23:15 | V1003 | 6.7310% | 6.8357% ↑ | 1.0153% | 1.0197% ↑ | 277.3% | 44,873 | ⏭️ Skip |
| 2026-10-11 20:15 | V1003 | 6.6646% | 6.7139% ↑ | 1.0141% | 1.0173% ↑ | 236.4% | 44,836 | ⏭️ Skip |
| 2026-10-11 17:14 | V1003 | 6.6024% | 6.6641% ↑ | 1.0105% | 1.0135% ↑ | 349.7% | 44,800 | ⏭️ Skip |
| 2026-10-11 14:13 | V1003 | 6.4841% | 6.5668% ↑ | 1.0106% | 1.0097% ↓ | 343.6% | 44,764 | ✅ Deploy |
| 2026-10-11 11:13 | V1003 | 6.3631% | 6.4390% ↑ | 1.0103% | 1.0111% ↑ | 268.9% | 44,728 | ⏭️ Skip |
| 2026-10-11 08:12 | V1003 | 6.2695% | 6.3196% ↑ | 1.0069% | 1.0059% ↓ | 259.0% | 44,692 | ✅ Deploy |
| 2026-10-11 05:11 | V1003 | 6.1977% | 6.2530% ↑ | 1.0059% | 1.0083% ↑ | 265.3% | 44,656 | ⏭️ Skip |
| 2026-10-11 02:10 | V1003 | 6.1121% | 6.1811% ↑ | 1.0055% | 1.0065% ↑ | 308.2% | 44,620 | ⏭️ Skip |
| 2026-10-10 23:10 | V1003 | 6.0043% | 6.0288% ↑ | 1.0043% | 1.0050% ↑ | 271.0% | 44,584 | ⏭️ Skip |
| 2026-10-10 20:09 | V1003 | 5.9234% | 5.9444% ↑ | 1.0033% | 1.0044% ↑ | 323.7% | 44,547 | ⏭️ Skip |
| 2026-10-10 17:09 | V1003 | 5.7879% | 5.8149% ↑ | 1.0006% | 1.0015% ↑ | 221.2% | 44,511 | ⏭️ Skip |
| 2026-10-10 14:08 | V1003 | 5.7166% | 5.7380% ↑ | 1.0001% | 0.9996% ↓ | 261.5% | 44,475 | ✅ Deploy |
| 2026-10-10 11:07 | V1003 | 5.6284% | 5.6357% ↑ | 0.9992% | 0.9997% ↑ | 272.4% | 44,439 | ⏭️ Skip |
| 2026-10-10 08:07 | V1003 | 5.5438% | 5.5569% ↑ | 0.9935% | 0.9947% ↑ | 260.7% | 44,403 | ⏭️ Skip |
| 2026-10-10 05:06 | V1003 | 5.4565% | 5.4461% ↓ | 0.9923% | 0.9927% ↑ | 319.8% | 44,367 | ✅ Deploy |
| 2026-10-10 02:05 | V1003 | 5.3259% | 5.3378% ↑ | 0.9908% | 0.9900% ↓ | 261.1% | 44,331 | ✅ Deploy |
| 2026-10-09 23:04 | V1003 | 5.2308% | 5.2300% ↓ | 0.9886% | 0.9891% ↑ | 206.0% | 44,294 | ✅ Deploy |
| 2026-10-09 20:04 | V1003 | 5.1458% | 5.1534% ↑ | 0.9893% | 0.9898% ↑ | 269.7% | 44,258 | ⏭️ Skip |
| 2026-10-09 17:03 | V1003 | 5.0497% | 5.0423% ↓ | 0.9875% | 0.9882% ↑ | 293.1% | 44,222 | ✅ Deploy |
| 2026-10-09 14:03 | V1003 | 4.9290% | 4.9295% ↑ | 0.9867% | 0.9872% ↑ | 290.8% | 44,186 | ⏭️ Skip |
| 2026-10-09 11:02 | V1003 | 4.8104% | 4.8106% ↑ | 0.9832% | 0.9843% ↑ | 331.6% | 44,150 | ⏭️ Skip |
| 2026-10-09 08:02 | V1003 | 4.6414% | 4.6505% ↑ | 0.9808% | 0.9817% ↑ | 278.2% | 44,114 | ⏭️ Skip |
| 2026-10-09 05:01 | V1003 | 4.5034% | 4.4957% ↓ | 0.9772% | 0.9784% ↑ | 197.9% | 44,078 | ✅ Deploy |
| 2026-10-09 02:00 | V1003 | 4.4191% | 4.4235% ↑ | 0.9760% | 0.9763% ↑ | 236.9% | 44,042 | ⏭️ Skip |
| 2026-10-08 22:59 | V1003 | 4.3244% | 4.3065% ↓ | 0.9761% | 0.9774% ↑ | 210.1% | 44,005 | ✅ Deploy |
| 2026-10-08 19:58 | V1003 | 4.1513% | 4.1531% ↑ | 0.9656% | 0.9665% ↑ | 61.2% | 43,969 | ⏭️ Skip |
| 2026-10-08 16:58 | V1003 | 4.1112% | 4.1150% ↑ | 0.9638% | 0.9653% ↑ | 46.7% | 43,933 | ⏭️ Skip |
| 2026-10-08 13:57 | V1003 | 4.0829% | 4.0868% ↑ | 0.9641% | 0.9621% ↓ | 49.9% | 43,897 | ✅ Deploy |
| 2026-10-08 10:56 | V1003 | 4.0537% | 4.0583% ↑ | 0.9633% | 0.9633% ↓ | 30.2% | 43,861 | ✅ Deploy |
| 2026-10-08 07:55 | V1003 | 4.0553% | 4.0602% ↑ | 0.9633% | 0.9638% ↑ | 41.8% | 43,825 | ⏭️ Skip |
| 2026-10-08 04:55 | V1003 | 4.0452% | 4.0475% ↑ | 0.9601% | 0.9599% ↓ | 15.5% | 43,788 | ✅ Deploy |
| 2026-10-08 01:54 | V1003 | 4.0626% | 4.0658% ↑ | 0.9590% | 0.9589% ↓ | 82.2% | 43,752 | ✅ Deploy |
| 2026-10-07 22:53 | V1003 | 4.0346% | 4.0387% ↑ | 0.9580% | 0.9572% ↓ | 93.8% | 43,716 | ✅ Deploy |
| 2026-10-07 19:52 | V1003 | 4.0017% | 4.0073% ↑ | 0.9522% | 0.9527% ↑ | 30.8% | 43,680 | ⏭️ Skip |
| 2026-10-07 16:52 | V1003 | 3.9985% | 4.0003% ↑ | 0.9473% | 0.9473% ↓ | 101.5% | 43,644 | ✅ Deploy |
| 2026-10-07 13:51 | V1003 | 3.9592% | 3.9647% ↑ | 0.9422% | 0.9422% ↓ | 81.1% | 43,608 | ✅ Deploy |
| 2026-10-07 10:50 | V1003 | 3.9341% | 3.9367% ↑ | 0.9415% | 0.9419% ↑ | 76.9% | 43,572 | ⏭️ Skip |
| 2026-10-07 07:49 | V1003 | 3.9142% | 3.9201% ↑ | 0.9384% | 0.9398% ↑ | 85.4% | 43,535 | ⏭️ Skip |
| 2026-10-07 04:49 | V1003 | 3.8821% | 3.8851% ↑ | 0.9368% | 0.9372% ↑ | 67.4% | 43,499 | ⏭️ Skip |
| 2026-10-07 01:48 | V1003 | 3.8642% | 3.8625% ↓ | 0.9319% | 0.9319% ↓ | 46.7% | 43,463 | ✅ Deploy |
| 2026-10-06 22:47 | V1003 | 3.8629% | 3.8623% ↓ | 0.9293% | 0.9303% ↑ | 72.4% | 43,427 | ✅ Deploy |
| 2026-10-06 19:46 | V1003 | 3.8458% | 3.8476% ↑ | 0.9288% | 0.9285% ↓ | 63.8% | 43,391 | ✅ Deploy |
| 2026-10-06 16:45 | V1003 | 3.8284% | 3.8309% ↑ | 0.9212% | 0.9214% ↑ | 86.0% | 43,355 | ⏭️ Skip |
কীভাবে কাজ করে
📊
ডেটা সংগ্রহ
প্রতি ৩০ সেকেন্ডে নতুন metrics
📈
Drift শনাক্ত
নতুন vs পুরনো mean (threshold 10%)
🔄
Retrain
XGBoost নতুন ডেটা দিয়ে train
✅
যাচাই
পুরনো vs নতুন RMSE তুলনা
🚀
Deploy
উন্নত হলেই replace