thanatchon36 commited on
Commit
103992e
·
verified ·
1 Parent(s): e25d050

Add model card, model specs, indicator config and the synthetic live nowcast

Browse files
Files changed (5) hide show
  1. LICENSE +21 -0
  2. README.md +142 -0
  3. config/indicators.yaml +319 -0
  4. live_nowcast.json +490 -0
  5. model_specs.json +387 -0
LICENSE ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ MIT License
2
+
3
+ Copyright (c) 2026 Thanatchon Chongmankhong
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
README.md ADDED
@@ -0,0 +1,142 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ - th
5
+ license: mit
6
+ library_name: scikit-learn
7
+ pipeline_tag: time-series-forecasting
8
+ datasets:
9
+ - thanatchon36/thainowcast
10
+ tags:
11
+ - nowcasting
12
+ - thailand
13
+ - macroeconomics
14
+ - gdp
15
+ - inflation
16
+ - dynamic-factor-model
17
+ - midas
18
+ - elastic-net
19
+ - lightgbm
20
+ - synthetic
21
+ ---
22
+
23
+ # thainowcast: nowcasting models for Thai GDP and inflation
24
+
25
+ > **No weights here, and no results about Thailand.** These models keep no fitted state. At every forecast
26
+ > origin each one is re-estimated from scratch on the data released by that date, so a frozen set of weights
27
+ > would be stale the day after it was saved. This repository records the models instead: the classes,
28
+ > hyperparameters, feature sets and the model × feature-set grid. Every number quoted below comes from
29
+ > **synthetic test data**. It shows that the pipeline works, not how well these models nowcast the Thai economy.
30
+
31
+ These are the models of the working paper *ThaiNowcast: Pseudo-Real-Time Nowcasting of Thai GDP and Inflation
32
+ with Official and Alternative Data* by Thanatchon Chongmankhong
33
+ ([read online](https://thanatchon36.github.io/thainowcast/paper/) ·
34
+ [PDF](https://github.com/thanatchon36/thainowcast/raw/main/paper/paper.pdf) ·
35
+ [DOI 10.5281/zenodo.23995547](https://doi.org/10.5281/zenodo.23995547)).
36
+ Code: <https://github.com/thanatchon36/thainowcast>. Data and full results:
37
+ [`thanatchon36/thainowcast`](https://huggingface.co/datasets/thanatchon36/thainowcast) (dataset).
38
+
39
+ *Independent personal research. Does not represent the views of any employer or institution.*
40
+
41
+ ## Targets
42
+
43
+ | target | frequency | forecast origins |
44
+ |---|---|---|
45
+ | `gdp_yoy`: real GDP growth, y-o-y (%), NESDC | quarterly | M1, M2, M3 (end of each month of the quarter), PR (the day before the official release) |
46
+ | `cpi_yoy`: headline CPI inflation, y-o-y (%), TPSO | monthly | M0 (mid-month), PR |
47
+
48
+ ## Models
49
+
50
+ | code | class | hyperparameters |
51
+ |---|---|---|
52
+ | `rw` | `RandomWalk` | last released value |
53
+ | `ar` | `AR` | AR(p), p ≤ 4 chosen by AIC |
54
+ | `bridge` | `Bridge` | indicators completed to the end of the target period with AR(1) forecasts, averaged to the target frequency, 2 principal components, OLS |
55
+ | `midas` | `MIDAS` | one regressor (mean of standardized indicators), 12 monthly lags, exponential Almon weights re-estimated for each number of missing months. GDP only |
56
+ | `dfm` | `DFM` | mixed-frequency dynamic factor model (statsmodels `DynamicFactorMQ`), 1 factor, AR(1) factor, EM up to 50 iterations, re-estimated once a year; also gives the news decomposition |
57
+ | `enet` | `ElasticNetNowcast` | `ElasticNetCV`, l1 ratio in {0.1, 0.5, 0.9}, 40 alphas, time-series CV with up to 4 folds |
58
+ | `lgbm` | `LightGBMNowcast` | 300 trees, learning rate 0.03, 4 leaves, min 5 samples per leaf, row and column subsampling 0.8, L2 1.0, deterministic |
59
+
60
+ Feature sets: `hard_only` (14 official monthly statistics), `hard_trends`, `hard_lights`, `hard_no2` and
61
+ `hard_prices` (hard data plus one family of alternative data: Google Trends, nighttime lights, Sentinel-5P NO2,
62
+ online prices), `all` (hard data plus all five families, mobility included) and `none` for the univariate
63
+ benchmarks. GDP runs 20 model × feature-set pairs and CPI 22. Each model's 80% interval is the conformal quantile
64
+ of its absolute backtest errors at the same origin.
65
+
66
+ ## Files
67
+
68
+ | path | content |
69
+ |---|---|
70
+ | `model_specs.json` | every model's class and hyperparameters, the indicators in each feature set, the backtest settings and the full grid per target, pinned to code commit `ce9dd21` |
71
+ | `config/indicators.yaml` | the indicator registry: agency, frequency, start, publication lag, transform and the source of each series |
72
+ | `live_nowcast.json` | **synthetic.** The live nowcast as of 2026-11-05: every model's value and interval, the headline model and the GDP news decomposition. Rerunning the snippet under Use reproduces the GDP headline, 2.622 |
73
+ | `LICENSE` | MIT, as for the code |
74
+
75
+ ## Accuracy on synthetic data
76
+
77
+ RMSFE at the last origin (PR) over the full backtest, relative to the AR benchmark (below 1 beats AR).
78
+ GDP covers 2015Q1 to 2026Q2 (46 quarters), CPI 2015-01 to 2026-09 (141 months).
79
+
80
+ | model | GDP `hard_only` | GDP `all` | CPI `hard_only` | CPI `all` |
81
+ |---|---|---|---|---|
82
+ | `enet` | **0.46** | 0.78 | 0.68 | **0.54** |
83
+ | `bridge` | 0.55 | 0.66 | 1.45 | 1.29 |
84
+ | `dfm` | 0.64 | 0.89 | 0.92 | 0.92 |
85
+ | `midas` | 0.64 | 0.66 | – | – |
86
+ | `lgbm` | 0.73 | 0.72 | 0.72 | 0.60 |
87
+ | `rw` | 1.04 | | 1.02 | |
88
+
89
+ The synthetic generator sets how much each data family helps, so this ranking tells you nothing about Thailand.
90
+ The full tables, Diebold-Mariano tests, regime splits and ablations are in the
91
+ [dataset](https://huggingface.co/datasets/thanatchon36/thainowcast/tree/main/results).
92
+
93
+ ## Use
94
+
95
+ The models run inside the pipeline, so you use the code rather than a download from this repository:
96
+
97
+ ```bash
98
+ git clone https://github.com/thanatchon36/thainowcast && cd thainowcast
99
+ make setup # uv venv + pip install -e ".[dev,app]"
100
+ make all # data -> eval (backtest) -> live -> dashboard -> paper
101
+ ```
102
+
103
+ To nowcast for real, put the official CSVs named in
104
+ [`docs/data_sources.md`](https://github.com/thanatchon36/thainowcast/blob/main/docs/data_sources.md) into
105
+ `data/raw/` (the `synthetic_*` files are used only when a real one is missing, with a warning), then run
106
+ `make all ASOF=YYYY-MM-DD`. One model on its own:
107
+
108
+ ```python
109
+ from thainowcast.config import load_config
110
+ from thainowcast.data.loaders import load_panel
111
+ from thainowcast.features import transform_wide
112
+ from thainowcast.models import MODELS
113
+ from thainowcast.config import RAW_DIR
114
+ import pandas as pd
115
+
116
+ cfg, asof = load_config(), pd.Timestamp("2026-11-05")
117
+ panel = load_panel(cfg, RAW_DIR)
118
+ y = panel.series_available(asof, "gdp_yoy") # GDP released by asof
119
+ X = transform_wide(panel.wide_available(asof, list(cfg.indicators), "M"), cfg.indicators)
120
+ X = X[cfg.set_columns("hard_only")]
121
+ print(MODELS["enet"]().nowcast(y, X, pd.Period("2026Q3", "Q"), asof))
122
+ ```
123
+
124
+ ## Limitations
125
+
126
+ - Trained and evaluated on synthetic data only. Real Thai series have revisions, breaks and gaps the generator
127
+ does not reproduce, and the backtest uses first releases, not real-time vintages.
128
+ - Publication lags in `config/indicators.yaml` are approximations of each agency's usual timing.
129
+ - The panel is short: 46 GDP quarters in the backtest, so differences between models are noisy.
130
+
131
+ ## Citation
132
+
133
+ ```bibtex
134
+ @misc{chongmankhong2026thainowcast,
135
+ author = {Chongmankhong, Thanatchon},
136
+ title = {ThaiNowcast: Pseudo-Real-Time Nowcasting of Thai GDP and Inflation with Official and Alternative Data},
137
+ year = {2026},
138
+ doi = {10.5281/zenodo.23995547},
139
+ url = {https://thanatchon36.github.io/thainowcast/paper/},
140
+ note = {Working paper; results are from synthetic test data}
141
+ }
142
+ ```
config/indicators.yaml ADDED
@@ -0,0 +1,319 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ThaiNowcast — indicator registry and release calendar
2
+ # ---------------------------------------------------------------------------
3
+ # start : first period of the LEVEL series (a YoY feature needs 12 more months)
4
+ # end : last period, only for discontinued series
5
+ # release_lag_days : typical calendar days between the end of the reference period and
6
+ # publication. Approximations taken from each agency's usual release
7
+ # timing — verify against the official calendars before a real run [ต้องตรวจ].
8
+ # transform : yoy | log_yoy | yoy_diff | level
9
+ # file : CSV expected in data/raw/ with columns `period,value`.
10
+ # `synthetic_<file>` is used ONLY when the real file is missing, with a warning.
11
+ # fetch : where the real file comes from (see docs/data_sources.md)
12
+
13
+ targets:
14
+ gdp_yoy:
15
+ name_th: "GDP จริง YoY (%)"
16
+ name_en: "Real GDP growth, y-o-y (%)"
17
+ agency: "NESDC (สภาพัฒน์)"
18
+ frequency: Q
19
+ start: "2005Q1"
20
+ release_lag_days: 48
21
+ transform: level
22
+ fetch: "manual: nesdc.go.th > QGDP press release > Excel table, YoY % of GDP at chained volume"
23
+ file: gdp_yoy.csv
24
+ gdp_qoq_sa:
25
+ name_th: "GDP จริง QoQ SA (%)"
26
+ name_en: "Real GDP growth, q-o-q seasonally adjusted (%)"
27
+ agency: "NESDC (สภาพัฒน์)"
28
+ frequency: Q
29
+ start: "2005Q1"
30
+ release_lag_days: 48
31
+ transform: level
32
+ fetch: "manual: same NESDC release as gdp_yoy, seasonally adjusted table"
33
+ file: gdp_qoq_sa.csv
34
+ cpi_yoy:
35
+ name_th: "เงินเฟ้อทั่วไป YoY (%)"
36
+ name_en: "Headline CPI inflation, y-o-y (%)"
37
+ agency: "TPSO กระทรวงพาณิชย์"
38
+ frequency: M
39
+ start: "2005-01"
40
+ release_lag_days: 6
41
+ transform: level
42
+ fetch: "manual: tpso.go.th / price.moc.go.th > CPI monthly report > headline YoY"
43
+ file: cpi_yoy.csv
44
+
45
+ indicators:
46
+ # ---------------- hard data: official monthly statistics ----------------
47
+ mpi:
48
+ name_th: "ดัชนีผลผลิตอุตสาหกรรม"
49
+ name_en: "Manufacturing Production Index"
50
+ agency: "OIE (สศอ.)"
51
+ group: hard
52
+ frequency: M
53
+ start: "2004-01"
54
+ release_lag_days: 30
55
+ transform: yoy
56
+ fetch: "manual: oie.go.th > สถิติอุตสาหกรรม > MPI xlsx (or BOT API EC_EI_027) [ต้องตรวจ]"
57
+ file: mpi.csv
58
+ pci:
59
+ name_th: "ดัชนีการบริโภคภาคเอกชน (PCI)"
60
+ name_en: "Private Consumption Index"
61
+ agency: "BOT (ธปท.)"
62
+ group: hard
63
+ frequency: M
64
+ start: "2004-01"
65
+ release_lag_days: 30
66
+ transform: yoy
67
+ fetch: "api: BOT API portal (apiportal.bot.or.th), Economic & Financial statistics, PCI [ต้องตรวจ series code]"
68
+ file: pci.csv
69
+ pii:
70
+ name_th: "ดัชนีการลงทุนภาคเอกชน (PII)"
71
+ name_en: "Private Investment Index"
72
+ agency: "BOT (ธปท.)"
73
+ group: hard
74
+ frequency: M
75
+ start: "2004-01"
76
+ release_lag_days: 30
77
+ transform: yoy
78
+ fetch: "api: BOT API portal, PII [ต้องตรวจ series code]"
79
+ file: pii.csv
80
+ credit:
81
+ name_th: "สินเชื่อระบบธนาคารพาณิชย์"
82
+ name_en: "Commercial bank loans outstanding"
83
+ agency: "BOT (ธปท.)"
84
+ group: hard
85
+ frequency: M
86
+ start: "2004-01"
87
+ release_lag_days: 30
88
+ transform: yoy
89
+ fetch: "api: BOT API portal, loans outstanding by sector"
90
+ file: credit.csv
91
+ promptpay:
92
+ name_th: "มูลค่าธุรกรรมพร้อมเพย์"
93
+ name_en: "PromptPay transaction value"
94
+ agency: "BOT (ธปท.)"
95
+ group: hard
96
+ frequency: M
97
+ start: "2016-01"
98
+ release_lag_days: 30
99
+ transform: yoy
100
+ fetch: "manual: BOT payment systems statistics (bot.or.th > Payment Systems > Statistics)"
101
+ file: promptpay.csv
102
+ exports:
103
+ name_th: "มูลค่าส่งออก (USD)"
104
+ name_en: "Merchandise exports (USD)"
105
+ agency: "Customs Department (กรมศุลกากร)"
106
+ group: hard
107
+ frequency: M
108
+ start: "2004-01"
109
+ release_lag_days: 25
110
+ transform: yoy
111
+ fetch: "manual: customs.go.th > สถิติการค้า; also MOC press release"
112
+ file: exports.csv
113
+ imports:
114
+ name_th: "มูลค่านำเข้า (USD)"
115
+ name_en: "Merchandise imports (USD)"
116
+ agency: "Customs Department (กรมศุลกากร)"
117
+ group: hard
118
+ frequency: M
119
+ start: "2004-01"
120
+ release_lag_days: 25
121
+ transform: yoy
122
+ fetch: "manual: customs.go.th > สถิติการค้า"
123
+ file: imports.csv
124
+ tourists:
125
+ name_th: "นักท่องเที่ยวต่างชาติ (คน)"
126
+ name_en: "International tourist arrivals"
127
+ agency: "MOTS (กระทรวงการท่องเที่ยวฯ)"
128
+ group: hard
129
+ frequency: M
130
+ start: "2004-01"
131
+ release_lag_days: 20
132
+ transform: log_yoy
133
+ fetch: "manual: mots.go.th > สถิตินักท่องเที่ยว > monthly arrivals"
134
+ file: tourists.csv
135
+ aot_pax:
136
+ name_th: "ผู้โดยสารสนามบิน AOT"
137
+ name_en: "AOT airport passengers"
138
+ agency: "AOT (ทอท.)"
139
+ group: hard
140
+ frequency: M
141
+ start: "2004-01"
142
+ release_lag_days: 20
143
+ transform: log_yoy
144
+ fetch: "manual: airportthai.co.th > Air Traffic Statistics"
145
+ file: aot_pax.csv
146
+ electricity:
147
+ name_th: "ปริมาณการใช้ไฟฟ้า"
148
+ name_en: "Electricity consumption"
149
+ agency: "EPPO (สนพ.)"
150
+ group: hard
151
+ frequency: M
152
+ start: "2004-01"
153
+ release_lag_days: 45
154
+ transform: yoy
155
+ fetch: "manual: eppo.go.th > สถิติพลังงาน > electricity consumption by sector"
156
+ file: electricity.csv
157
+ vehicle_reg:
158
+ name_th: "ยอดจดทะเบียนรถใหม่"
159
+ name_en: "New vehicle registrations"
160
+ agency: "DLT (กรมการขนส่งทางบก)"
161
+ group: hard
162
+ frequency: M
163
+ start: "2004-01"
164
+ release_lag_days: 15
165
+ transform: yoy
166
+ fetch: "manual: web.dlt.go.th/statistics > จดทะเบียนใหม่"
167
+ file: vehicle_reg.csv
168
+ expressway:
169
+ name_th: "ปริมาณจราจรทางพิเศษ"
170
+ name_en: "Expressway traffic volume"
171
+ agency: "EXAT (กทพ.)"
172
+ group: hard
173
+ frequency: M
174
+ start: "2004-01"
175
+ release_lag_days: 30
176
+ transform: yoy
177
+ fetch: "manual: exat.co.th > สถิติปริมาณจราจร"
178
+ file: expressway.csv
179
+ oil_retail:
180
+ name_th: "ราคาขายปลีกดีเซล"
181
+ name_en: "Retail diesel price (monthly average)"
182
+ agency: "EPPO (สนพ.) / PTT"
183
+ group: hard
184
+ frequency: M
185
+ start: "2004-01"
186
+ release_lag_days: 1
187
+ transform: yoy
188
+ fetch: "scrape: eppo.go.th petroleum price statistics (daily, average to month)"
189
+ file: oil_retail.csv
190
+ agri_price:
191
+ name_th: "ดัชนีราคาสินค้าเกษตร"
192
+ name_en: "Farm price index"
193
+ agency: "OAE (สศก.)"
194
+ group: hard
195
+ frequency: M
196
+ start: "2004-01"
197
+ release_lag_days: 15
198
+ transform: yoy
199
+ fetch: "manual: oae.go.th > ดัชนีราคาสินค้าเกษตร"
200
+ file: agri_price.csv
201
+
202
+ # ---------------- alternative data ----------------
203
+ gt_unemploy:
204
+ name_th: "Google Trends: ว่างงาน / หางาน"
205
+ name_en: "Google Trends: unemployment / job search (TH)"
206
+ agency: "Google Trends"
207
+ group: alt_trends
208
+ frequency: M
209
+ start: "2008-01"
210
+ release_lag_days: 3
211
+ transform: yoy_diff
212
+ fetch: "api: pytrends, geo=TH, topic 'ว่างงาน','หางาน'; weekly -> monthly mean"
213
+ file: gt_unemploy.csv
214
+ gt_loan:
215
+ name_th: "Google Trends: ขอสินเชื่อ / กู้เงิน"
216
+ name_en: "Google Trends: loan / borrowing (TH)"
217
+ agency: "Google Trends"
218
+ group: alt_trends
219
+ frequency: M
220
+ start: "2008-01"
221
+ release_lag_days: 3
222
+ transform: yoy_diff
223
+ fetch: "api: pytrends, geo=TH, 'กู้เงิน','สินเชื่อ'"
224
+ file: gt_loan.csv
225
+ gt_car_installment:
226
+ name_th: "Google Trends: ผ่อนรถ"
227
+ name_en: "Google Trends: car installment (TH)"
228
+ agency: "Google Trends"
229
+ group: alt_trends
230
+ frequency: M
231
+ start: "2008-01"
232
+ release_lag_days: 3
233
+ transform: yoy_diff
234
+ fetch: "api: pytrends, geo=TH, 'ผ่อนรถ'"
235
+ file: gt_car_installment.csv
236
+ gt_travel:
237
+ name_th: "Google Trends: ท่องเที่ยว / จองโรงแรม"
238
+ name_en: "Google Trends: travel / hotel booking (TH)"
239
+ agency: "Google Trends"
240
+ group: alt_trends
241
+ frequency: M
242
+ start: "2008-01"
243
+ release_lag_days: 3
244
+ transform: yoy_diff
245
+ fetch: "api: pytrends, geo=TH, 'ที่พัก','จองโรงแรม'"
246
+ file: gt_travel.csv
247
+ gt_prices:
248
+ name_th: "Google Trends: ของแพง / ราคาน้ำมัน"
249
+ name_en: "Google Trends: prices / fuel price (TH)"
250
+ agency: "Google Trends"
251
+ group: alt_trends
252
+ frequency: M
253
+ start: "2008-01"
254
+ release_lag_days: 3
255
+ transform: yoy_diff
256
+ fetch: "api: pytrends, geo=TH, 'ของแพง','ราคาน้ำมันวันนี้'"
257
+ file: gt_prices.csv
258
+ nightlights:
259
+ name_th: "แสงไฟกลางคืน (VIIRS Black Marble)"
260
+ name_en: "Nighttime lights radiance, national sum (VNP46A3)"
261
+ agency: "NASA Black Marble"
262
+ group: alt_lights
263
+ frequency: M
264
+ start: "2012-04"
265
+ release_lag_days: 40
266
+ transform: log_yoy
267
+ fetch: "api: NASA Earthdata (ladsweb) VNP46A3 monthly tiles -> provincial zonal sum with GADM boundaries"
268
+ file: nightlights.csv
269
+ no2:
270
+ name_th: "NO2 โทรโพสเฟียร์ (Sentinel-5P)"
271
+ name_en: "Tropospheric NO2 column, industrial provinces (Sentinel-5P)"
272
+ agency: "ESA Copernicus"
273
+ group: alt_no2
274
+ frequency: M
275
+ start: "2018-07"
276
+ release_lag_days: 5
277
+ transform: log_yoy
278
+ fetch: "api: Google Earth Engine COPERNICUS/S5P/OFFL/L3_NO2, monthly mean over EEC + Samut Prakan"
279
+ file: no2.csv
280
+ online_prices:
281
+ name_th: "ดัชนีราคาสินค้าออนไลน์"
282
+ name_en: "Online price index (scraped daily, Billion-Prices style)"
283
+ agency: "own scraper"
284
+ group: alt_prices
285
+ frequency: M
286
+ start: "2015-01"
287
+ release_lag_days: 1
288
+ transform: yoy
289
+ fetch: "scrape: own daily scraper of major online grocers (check ToS) [ต้องตรวจ]"
290
+ file: online_prices.csv
291
+ mobility:
292
+ name_th: "Google Mobility: retail & recreation"
293
+ name_en: "Google Community Mobility: retail & recreation (% vs baseline)"
294
+ agency: "Google (discontinued 2022-10)"
295
+ group: alt_mobility
296
+ frequency: M
297
+ start: "2020-02"
298
+ end: "2022-10"
299
+ release_lag_days: 2
300
+ transform: level
301
+ fetch: "manual: google.com/covid19/mobility archived CSV, TH national"
302
+ file: mobility.csv
303
+
304
+ feature_sets:
305
+ hard_only: [hard]
306
+ hard_trends: [hard, alt_trends]
307
+ hard_lights: [hard, alt_lights]
308
+ hard_no2: [hard, alt_no2]
309
+ hard_prices: [hard, alt_prices]
310
+ all: [hard, alt_trends, alt_lights, alt_no2, alt_prices, alt_mobility]
311
+
312
+ backtest:
313
+ start: "2015-01"
314
+ benchmark: ar
315
+ conformal_alpha: 0.2
316
+ regimes:
317
+ pre_covid: ["2015-01", "2019-12"]
318
+ covid: ["2020-01", "2021-12"]
319
+ post_covid: ["2022-01", "2035-12"]
live_nowcast.json ADDED
@@ -0,0 +1,490 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "asof": "2026-11-05",
3
+ "synthetic": true,
4
+ "synthetic_ids": [
5
+ "agri_price",
6
+ "aot_pax",
7
+ "cpi_yoy",
8
+ "credit",
9
+ "electricity",
10
+ "exports",
11
+ "expressway",
12
+ "gdp_qoq_sa",
13
+ "gdp_yoy",
14
+ "gt_car_installment",
15
+ "gt_loan",
16
+ "gt_prices",
17
+ "gt_travel",
18
+ "gt_unemploy",
19
+ "imports",
20
+ "mobility",
21
+ "mpi",
22
+ "nightlights",
23
+ "no2",
24
+ "oil_retail",
25
+ "online_prices",
26
+ "pci",
27
+ "pii",
28
+ "promptpay",
29
+ "tourists",
30
+ "vehicle_reg"
31
+ ],
32
+ "targets": {
33
+ "gdp_yoy": {
34
+ "label": "Real GDP growth (y-o-y)",
35
+ "period": "2026Q3",
36
+ "release_expected": "2026-11-17",
37
+ "headline": {
38
+ "model": "enet",
39
+ "feature_set": "hard_only",
40
+ "value": 2.622081984570472,
41
+ "lo": 1.2321626043496732,
42
+ "hi": 4.012001364791271,
43
+ "rmsfe_full": 1.1729679933805253
44
+ },
45
+ "models": [
46
+ {
47
+ "model": "rw",
48
+ "feature_set": "none",
49
+ "value": 4.035735,
50
+ "lo": 2.1696509999999996,
51
+ "hi": 5.901819
52
+ },
53
+ {
54
+ "model": "ar",
55
+ "feature_set": "none",
56
+ "value": 3.7247021640278297,
57
+ "lo": 1.6627235019632787,
58
+ "hi": 5.786680826092381
59
+ },
60
+ {
61
+ "model": "bridge",
62
+ "feature_set": "hard_only",
63
+ "value": 2.744188700020815,
64
+ "lo": 1.1611176888087378,
65
+ "hi": 4.327259711232892
66
+ },
67
+ {
68
+ "model": "bridge",
69
+ "feature_set": "hard_trends",
70
+ "value": 2.674056418037423,
71
+ "lo": 1.2520382531462815,
72
+ "hi": 4.096074582928565
73
+ },
74
+ {
75
+ "model": "bridge",
76
+ "feature_set": "hard_lights",
77
+ "value": 2.771415556231245,
78
+ "lo": 1.43295525583927,
79
+ "hi": 4.109875856623219
80
+ },
81
+ {
82
+ "model": "bridge",
83
+ "feature_set": "hard_no2",
84
+ "value": 2.496059678850094,
85
+ "lo": 0.912988667638017,
86
+ "hi": 4.079130690062171
87
+ },
88
+ {
89
+ "model": "bridge",
90
+ "feature_set": "hard_prices",
91
+ "value": 2.62062591496925,
92
+ "lo": 1.1868895915257238,
93
+ "hi": 4.054362238412777
94
+ },
95
+ {
96
+ "model": "bridge",
97
+ "feature_set": "all",
98
+ "value": 2.4001047258769646,
99
+ "lo": 1.0819277773060407,
100
+ "hi": 3.7182816744478884
101
+ },
102
+ {
103
+ "model": "enet",
104
+ "feature_set": "hard_only",
105
+ "value": 2.622081984570472,
106
+ "lo": 1.2321626043496732,
107
+ "hi": 4.012001364791271
108
+ },
109
+ {
110
+ "model": "enet",
111
+ "feature_set": "hard_trends",
112
+ "value": 2.5062786974179816,
113
+ "lo": 1.1956021229098588,
114
+ "hi": 3.8169552719261044
115
+ },
116
+ {
117
+ "model": "enet",
118
+ "feature_set": "hard_lights",
119
+ "value": 2.622081984570472,
120
+ "lo": 1.2958890611390879,
121
+ "hi": 3.948274908001856
122
+ },
123
+ {
124
+ "model": "enet",
125
+ "feature_set": "hard_no2",
126
+ "value": 2.0694009258166295,
127
+ "lo": 0.6794815455958307,
128
+ "hi": 3.4593203060374282
129
+ },
130
+ {
131
+ "model": "enet",
132
+ "feature_set": "hard_prices",
133
+ "value": 2.6069274801172466,
134
+ "lo": 1.265976164598964,
135
+ "hi": 3.9478787956355292
136
+ },
137
+ {
138
+ "model": "enet",
139
+ "feature_set": "all",
140
+ "value": 2.06940092581663,
141
+ "lo": 0.7471238227565626,
142
+ "hi": 3.3916780288766972
143
+ },
144
+ {
145
+ "model": "midas",
146
+ "feature_set": "hard_only",
147
+ "value": 3.4239859732768743,
148
+ "lo": 1.3072362627081366,
149
+ "hi": 5.540735683845612
150
+ },
151
+ {
152
+ "model": "midas",
153
+ "feature_set": "all",
154
+ "value": 3.2150876254650664,
155
+ "lo": 1.1521296518849318,
156
+ "hi": 5.2780455990452015
157
+ },
158
+ {
159
+ "model": "lgbm",
160
+ "feature_set": "hard_only",
161
+ "value": 2.9411170887132343,
162
+ "lo": 1.3701709870669854,
163
+ "hi": 4.512063190359483
164
+ },
165
+ {
166
+ "model": "lgbm",
167
+ "feature_set": "all",
168
+ "value": 3.101451725661528,
169
+ "lo": 1.5555462561563083,
170
+ "hi": 4.647357195166748
171
+ },
172
+ {
173
+ "model": "dfm",
174
+ "feature_set": "hard_only",
175
+ "value": 3.6750213539314704,
176
+ "lo": 2.013620414584559,
177
+ "hi": 5.336422293278382
178
+ },
179
+ {
180
+ "model": "dfm",
181
+ "feature_set": "all",
182
+ "value": 3.390650768988231,
183
+ "lo": 1.781317465636148,
184
+ "hi": 4.999984072340314
185
+ }
186
+ ],
187
+ "origins": [
188
+ {
189
+ "horizon": "M1",
190
+ "origin": "2026-07-31"
191
+ },
192
+ {
193
+ "horizon": "M2",
194
+ "origin": "2026-08-31"
195
+ },
196
+ {
197
+ "horizon": "M3",
198
+ "origin": "2026-09-30"
199
+ },
200
+ {
201
+ "horizon": "latest",
202
+ "origin": "2026-11-05"
203
+ }
204
+ ]
205
+ },
206
+ "cpi_yoy": {
207
+ "label": "Headline CPI inflation (y-o-y)",
208
+ "period": "2026-10",
209
+ "release_expected": "2026-11-06",
210
+ "headline": {
211
+ "model": "enet",
212
+ "feature_set": "all",
213
+ "value": 2.3061274905583997,
214
+ "lo": 1.7569674731498859,
215
+ "hi": 2.8552875079669136,
216
+ "rmsfe_full": 0.4361894121007024
217
+ },
218
+ "models": [
219
+ {
220
+ "model": "rw",
221
+ "feature_set": "none",
222
+ "value": 1.986116,
223
+ "lo": 1.2255880000000001,
224
+ "hi": 2.746644
225
+ },
226
+ {
227
+ "model": "ar",
228
+ "feature_set": "none",
229
+ "value": 2.00337733475106,
230
+ "lo": 1.268319552720219,
231
+ "hi": 2.7384351167819014
232
+ },
233
+ {
234
+ "model": "bridge",
235
+ "feature_set": "hard_only",
236
+ "value": 2.0678123266897193,
237
+ "lo": 0.6540992818840159,
238
+ "hi": 3.4815253714954224
239
+ },
240
+ {
241
+ "model": "bridge",
242
+ "feature_set": "hard_trends",
243
+ "value": 1.9620034720025499,
244
+ "lo": 0.4891452396813707,
245
+ "hi": 3.434861704323729
246
+ },
247
+ {
248
+ "model": "bridge",
249
+ "feature_set": "hard_lights",
250
+ "value": 2.0765670156961114,
251
+ "lo": 0.37655284097829655,
252
+ "hi": 3.776581190413926
253
+ },
254
+ {
255
+ "model": "bridge",
256
+ "feature_set": "hard_no2",
257
+ "value": 1.9200449635415748,
258
+ "lo": 0.6043047433915709,
259
+ "hi": 3.2357851836915787
260
+ },
261
+ {
262
+ "model": "bridge",
263
+ "feature_set": "hard_prices",
264
+ "value": 2.2296769839880586,
265
+ "lo": 1.1140155832432992,
266
+ "hi": 3.345338384732818
267
+ },
268
+ {
269
+ "model": "bridge",
270
+ "feature_set": "all",
271
+ "value": 1.975565967816605,
272
+ "lo": 0.6607716191167214,
273
+ "hi": 3.2903603165164887
274
+ },
275
+ {
276
+ "model": "enet",
277
+ "feature_set": "hard_only",
278
+ "value": 1.4688316407918371,
279
+ "lo": 0.7576763015292372,
280
+ "hi": 2.179986980054437
281
+ },
282
+ {
283
+ "model": "enet",
284
+ "feature_set": "hard_trends",
285
+ "value": 1.5680678603109373,
286
+ "lo": 0.8654972705394863,
287
+ "hi": 2.270638450082388
288
+ },
289
+ {
290
+ "model": "enet",
291
+ "feature_set": "hard_lights",
292
+ "value": 1.453471697024606,
293
+ "lo": 0.7403797703699131,
294
+ "hi": 2.166563623679299
295
+ },
296
+ {
297
+ "model": "enet",
298
+ "feature_set": "hard_no2",
299
+ "value": 1.4918525949020902,
300
+ "lo": 0.7879344711507577,
301
+ "hi": 2.1957707186534225
302
+ },
303
+ {
304
+ "model": "enet",
305
+ "feature_set": "hard_prices",
306
+ "value": 2.3240561431880207,
307
+ "lo": 1.7618486314697233,
308
+ "hi": 2.886263654906318
309
+ },
310
+ {
311
+ "model": "enet",
312
+ "feature_set": "all",
313
+ "value": 2.3061274905583997,
314
+ "lo": 1.7569674731498859,
315
+ "hi": 2.8552875079669136
316
+ },
317
+ {
318
+ "model": "lgbm",
319
+ "feature_set": "hard_only",
320
+ "value": 1.7115373151574091,
321
+ "lo": 0.9853831938492812,
322
+ "hi": 2.437691436465537
323
+ },
324
+ {
325
+ "model": "lgbm",
326
+ "feature_set": "hard_trends",
327
+ "value": 1.8052224662652232,
328
+ "lo": 1.0741084133932657,
329
+ "hi": 2.5363365191371807
330
+ },
331
+ {
332
+ "model": "lgbm",
333
+ "feature_set": "hard_lights",
334
+ "value": 1.7343294825051507,
335
+ "lo": 0.9627479740720566,
336
+ "hi": 2.505910990938245
337
+ },
338
+ {
339
+ "model": "lgbm",
340
+ "feature_set": "hard_no2",
341
+ "value": 1.601669895932529,
342
+ "lo": 0.8755157746244011,
343
+ "hi": 2.327824017240657
344
+ },
345
+ {
346
+ "model": "lgbm",
347
+ "feature_set": "hard_prices",
348
+ "value": 2.6627841727515746,
349
+ "lo": 2.042484129772815,
350
+ "hi": 3.283084215730334
351
+ },
352
+ {
353
+ "model": "lgbm",
354
+ "feature_set": "all",
355
+ "value": 2.4313120467587654,
356
+ "lo": 1.8900278903158707,
357
+ "hi": 2.97259620320166
358
+ },
359
+ {
360
+ "model": "dfm",
361
+ "feature_set": "hard_only",
362
+ "value": 1.9848952359169132,
363
+ "lo": 1.2595816683075094,
364
+ "hi": 2.710208803526317
365
+ },
366
+ {
367
+ "model": "dfm",
368
+ "feature_set": "all",
369
+ "value": 1.9386500191541647,
370
+ "lo": 1.1784398057771612,
371
+ "hi": 2.6988602325311684
372
+ }
373
+ ],
374
+ "origins": [
375
+ {
376
+ "horizon": "mid",
377
+ "origin": "2026-10-15"
378
+ },
379
+ {
380
+ "horizon": "M0",
381
+ "origin": "2026-10-31"
382
+ },
383
+ {
384
+ "horizon": "PR",
385
+ "origin": "2026-11-05"
386
+ }
387
+ ]
388
+ }
389
+ },
390
+ "news": {
391
+ "target": "gdp_yoy",
392
+ "feature_set": "all",
393
+ "from": "2026-09-30",
394
+ "to": "2026-11-05",
395
+ "nowcast_prev": 3.5685982984504783,
396
+ "nowcast_new": 3.3940234709497727,
397
+ "change": -0.17457482750070552,
398
+ "method": "statsmodels news",
399
+ "by_series": [
400
+ {
401
+ "series": "gt_travel",
402
+ "impact": -0.1086785742316695
403
+ },
404
+ {
405
+ "series": "gt_unemploy",
406
+ "impact": -0.04016759006269474
407
+ },
408
+ {
409
+ "series": "vehicle_reg",
410
+ "impact": -0.036649334000346166
411
+ },
412
+ {
413
+ "series": "gt_loan",
414
+ "impact": -0.01628605938699763
415
+ },
416
+ {
417
+ "series": "imports",
418
+ "impact": -0.015136326294564404
419
+ },
420
+ {
421
+ "series": "expressway",
422
+ "impact": -0.009875545780666455
423
+ },
424
+ {
425
+ "series": "electricity",
426
+ "impact": -0.0078903737892733
427
+ },
428
+ {
429
+ "series": "aot_pax",
430
+ "impact": -0.0030941998220819413
431
+ },
432
+ {
433
+ "series": "gt_car_installment",
434
+ "impact": -0.0018230290502531492
435
+ },
436
+ {
437
+ "series": "tourists",
438
+ "impact": -0.00153754350423599
439
+ },
440
+ {
441
+ "series": "gt_prices",
442
+ "impact": -0.000554228192775243
443
+ },
444
+ {
445
+ "series": "agri_price",
446
+ "impact": -0.00011514053254522077
447
+ },
448
+ {
449
+ "series": "no2",
450
+ "impact": 8.707749035095566e-05
451
+ },
452
+ {
453
+ "series": "oil_retail",
454
+ "impact": 0.0003556778514359496
455
+ },
456
+ {
457
+ "series": "promptpay",
458
+ "impact": 0.00046628805263241923
459
+ },
460
+ {
461
+ "series": "pii",
462
+ "impact": 0.0012703540228690478
463
+ },
464
+ {
465
+ "series": "online_prices",
466
+ "impact": 0.0017428370827895715
467
+ },
468
+ {
469
+ "series": "mpi",
470
+ "impact": 0.001969272432995771
471
+ },
472
+ {
473
+ "series": "credit",
474
+ "impact": 0.002661494191363973
475
+ },
476
+ {
477
+ "series": "exports",
478
+ "impact": 0.010002031045984879
479
+ },
480
+ {
481
+ "series": "nightlights",
482
+ "impact": 0.014907920357179036
483
+ },
484
+ {
485
+ "series": "pci",
486
+ "impact": 0.0300623035927379
487
+ }
488
+ ]
489
+ }
490
+ }
model_specs.json ADDED
@@ -0,0 +1,387 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "package": "thainowcast",
3
+ "version": "0.1.0",
4
+ "code": "https://github.com/thanatchon36/thainowcast",
5
+ "commit": "ce9dd21fafbf9adcd649900c8a006a776bdc6d4c",
6
+ "trained_on": "synthetic test data (seed 20261105); no fitted weights are stored, every model is re-estimated at each origin from the data released by that date",
7
+ "models": {
8
+ "rw": {
9
+ "class": "thainowcast.models.benchmarks.RandomWalk",
10
+ "params": {}
11
+ },
12
+ "ar": {
13
+ "class": "thainowcast.models.benchmarks.AR",
14
+ "params": {
15
+ "max_lags": 4
16
+ }
17
+ },
18
+ "bridge": {
19
+ "class": "thainowcast.models.bridge.Bridge",
20
+ "params": {
21
+ "n_factors": 2
22
+ }
23
+ },
24
+ "midas": {
25
+ "class": "thainowcast.models.midas.MIDAS",
26
+ "params": {
27
+ "n_lags": 12
28
+ }
29
+ },
30
+ "dfm": {
31
+ "class": "thainowcast.models.dfm.DFM",
32
+ "params": {
33
+ "n_factors": 1,
34
+ "factor_order": 1,
35
+ "maxiter": 50,
36
+ "refit": "annual"
37
+ }
38
+ },
39
+ "enet": {
40
+ "class": "thainowcast.models.ml.ElasticNetNowcast",
41
+ "params": {
42
+ "l1_ratios": [
43
+ 0.1,
44
+ 0.5,
45
+ 0.9
46
+ ],
47
+ "n_splits": 4
48
+ }
49
+ },
50
+ "lgbm": {
51
+ "class": "thainowcast.models.ml.LightGBMNowcast",
52
+ "params": {
53
+ "n_estimators": 300,
54
+ "learning_rate": 0.03,
55
+ "num_leaves": 4,
56
+ "min_child_samples": 5
57
+ }
58
+ }
59
+ },
60
+ "feature_sets": {
61
+ "hard_only": [
62
+ "mpi",
63
+ "pci",
64
+ "pii",
65
+ "credit",
66
+ "promptpay",
67
+ "exports",
68
+ "imports",
69
+ "tourists",
70
+ "aot_pax",
71
+ "electricity",
72
+ "vehicle_reg",
73
+ "expressway",
74
+ "oil_retail",
75
+ "agri_price"
76
+ ],
77
+ "hard_trends": [
78
+ "mpi",
79
+ "pci",
80
+ "pii",
81
+ "credit",
82
+ "promptpay",
83
+ "exports",
84
+ "imports",
85
+ "tourists",
86
+ "aot_pax",
87
+ "electricity",
88
+ "vehicle_reg",
89
+ "expressway",
90
+ "oil_retail",
91
+ "agri_price",
92
+ "gt_unemploy",
93
+ "gt_loan",
94
+ "gt_car_installment",
95
+ "gt_travel",
96
+ "gt_prices"
97
+ ],
98
+ "hard_lights": [
99
+ "mpi",
100
+ "pci",
101
+ "pii",
102
+ "credit",
103
+ "promptpay",
104
+ "exports",
105
+ "imports",
106
+ "tourists",
107
+ "aot_pax",
108
+ "electricity",
109
+ "vehicle_reg",
110
+ "expressway",
111
+ "oil_retail",
112
+ "agri_price",
113
+ "nightlights"
114
+ ],
115
+ "hard_no2": [
116
+ "mpi",
117
+ "pci",
118
+ "pii",
119
+ "credit",
120
+ "promptpay",
121
+ "exports",
122
+ "imports",
123
+ "tourists",
124
+ "aot_pax",
125
+ "electricity",
126
+ "vehicle_reg",
127
+ "expressway",
128
+ "oil_retail",
129
+ "agri_price",
130
+ "no2"
131
+ ],
132
+ "hard_prices": [
133
+ "mpi",
134
+ "pci",
135
+ "pii",
136
+ "credit",
137
+ "promptpay",
138
+ "exports",
139
+ "imports",
140
+ "tourists",
141
+ "aot_pax",
142
+ "electricity",
143
+ "vehicle_reg",
144
+ "expressway",
145
+ "oil_retail",
146
+ "agri_price",
147
+ "online_prices"
148
+ ],
149
+ "all": [
150
+ "mpi",
151
+ "pci",
152
+ "pii",
153
+ "credit",
154
+ "promptpay",
155
+ "exports",
156
+ "imports",
157
+ "tourists",
158
+ "aot_pax",
159
+ "electricity",
160
+ "vehicle_reg",
161
+ "expressway",
162
+ "oil_retail",
163
+ "agri_price",
164
+ "gt_unemploy",
165
+ "gt_loan",
166
+ "gt_car_installment",
167
+ "gt_travel",
168
+ "gt_prices",
169
+ "nightlights",
170
+ "no2",
171
+ "online_prices",
172
+ "mobility"
173
+ ],
174
+ "none": []
175
+ },
176
+ "backtest": {
177
+ "start": "2015-01",
178
+ "benchmark": "ar",
179
+ "conformal_alpha": 0.2,
180
+ "regimes": {
181
+ "pre_covid": [
182
+ "2015-01",
183
+ "2019-12"
184
+ ],
185
+ "covid": [
186
+ "2020-01",
187
+ "2021-12"
188
+ ],
189
+ "post_covid": [
190
+ "2022-01",
191
+ "2035-12"
192
+ ]
193
+ }
194
+ },
195
+ "targets": {
196
+ "gdp_yoy": {
197
+ "frequency": "Q",
198
+ "origins": [
199
+ "M1",
200
+ "M2",
201
+ "M3",
202
+ "PR"
203
+ ],
204
+ "final_origin": "PR",
205
+ "grid": [
206
+ {
207
+ "model": "rw",
208
+ "feature_set": "none"
209
+ },
210
+ {
211
+ "model": "ar",
212
+ "feature_set": "none"
213
+ },
214
+ {
215
+ "model": "bridge",
216
+ "feature_set": "hard_only"
217
+ },
218
+ {
219
+ "model": "bridge",
220
+ "feature_set": "hard_trends"
221
+ },
222
+ {
223
+ "model": "bridge",
224
+ "feature_set": "hard_lights"
225
+ },
226
+ {
227
+ "model": "bridge",
228
+ "feature_set": "hard_no2"
229
+ },
230
+ {
231
+ "model": "bridge",
232
+ "feature_set": "hard_prices"
233
+ },
234
+ {
235
+ "model": "bridge",
236
+ "feature_set": "all"
237
+ },
238
+ {
239
+ "model": "enet",
240
+ "feature_set": "hard_only"
241
+ },
242
+ {
243
+ "model": "enet",
244
+ "feature_set": "hard_trends"
245
+ },
246
+ {
247
+ "model": "enet",
248
+ "feature_set": "hard_lights"
249
+ },
250
+ {
251
+ "model": "enet",
252
+ "feature_set": "hard_no2"
253
+ },
254
+ {
255
+ "model": "enet",
256
+ "feature_set": "hard_prices"
257
+ },
258
+ {
259
+ "model": "enet",
260
+ "feature_set": "all"
261
+ },
262
+ {
263
+ "model": "midas",
264
+ "feature_set": "hard_only"
265
+ },
266
+ {
267
+ "model": "midas",
268
+ "feature_set": "all"
269
+ },
270
+ {
271
+ "model": "lgbm",
272
+ "feature_set": "hard_only"
273
+ },
274
+ {
275
+ "model": "lgbm",
276
+ "feature_set": "all"
277
+ },
278
+ {
279
+ "model": "dfm",
280
+ "feature_set": "hard_only"
281
+ },
282
+ {
283
+ "model": "dfm",
284
+ "feature_set": "all"
285
+ }
286
+ ]
287
+ },
288
+ "cpi_yoy": {
289
+ "frequency": "M",
290
+ "origins": [
291
+ "M0",
292
+ "PR"
293
+ ],
294
+ "final_origin": "PR",
295
+ "grid": [
296
+ {
297
+ "model": "rw",
298
+ "feature_set": "none"
299
+ },
300
+ {
301
+ "model": "ar",
302
+ "feature_set": "none"
303
+ },
304
+ {
305
+ "model": "bridge",
306
+ "feature_set": "hard_only"
307
+ },
308
+ {
309
+ "model": "bridge",
310
+ "feature_set": "hard_trends"
311
+ },
312
+ {
313
+ "model": "bridge",
314
+ "feature_set": "hard_lights"
315
+ },
316
+ {
317
+ "model": "bridge",
318
+ "feature_set": "hard_no2"
319
+ },
320
+ {
321
+ "model": "bridge",
322
+ "feature_set": "hard_prices"
323
+ },
324
+ {
325
+ "model": "bridge",
326
+ "feature_set": "all"
327
+ },
328
+ {
329
+ "model": "enet",
330
+ "feature_set": "hard_only"
331
+ },
332
+ {
333
+ "model": "enet",
334
+ "feature_set": "hard_trends"
335
+ },
336
+ {
337
+ "model": "enet",
338
+ "feature_set": "hard_lights"
339
+ },
340
+ {
341
+ "model": "enet",
342
+ "feature_set": "hard_no2"
343
+ },
344
+ {
345
+ "model": "enet",
346
+ "feature_set": "hard_prices"
347
+ },
348
+ {
349
+ "model": "enet",
350
+ "feature_set": "all"
351
+ },
352
+ {
353
+ "model": "lgbm",
354
+ "feature_set": "hard_only"
355
+ },
356
+ {
357
+ "model": "lgbm",
358
+ "feature_set": "hard_trends"
359
+ },
360
+ {
361
+ "model": "lgbm",
362
+ "feature_set": "hard_lights"
363
+ },
364
+ {
365
+ "model": "lgbm",
366
+ "feature_set": "hard_no2"
367
+ },
368
+ {
369
+ "model": "lgbm",
370
+ "feature_set": "hard_prices"
371
+ },
372
+ {
373
+ "model": "lgbm",
374
+ "feature_set": "all"
375
+ },
376
+ {
377
+ "model": "dfm",
378
+ "feature_set": "hard_only"
379
+ },
380
+ {
381
+ "model": "dfm",
382
+ "feature_set": "all"
383
+ }
384
+ ]
385
+ }
386
+ }
387
+ }