tanml 0.1.6__py3-none-any.whl → 0.1.7__py3-none-any.whl

tanml/utils/data_loader.py

@@ -1,17 +1,107 @@
-import os
+from __future__ import annotations
+
+from pathlib import Path
+from typing import Any, Optional
 import pandas as pd
 
-def load_dataframe(filepath):
-    ext = os.path.splitext(filepath)[1].lower()
-    if ext == ".csv":
-        return pd.read_csv(filepath)
-    elif ext in [".xls", ".xlsx"]:
-        return pd.read_excel(filepath)
-    elif ext == ".parquet":
-        return pd.read_parquet(filepath)
-    elif ext == ".sas7bdat":
-        return pd.read_sas(filepath)
-    elif ext in [".txt", ".tsv"]:
-        return pd.read_csv(filepath, sep="\t")
-    else:
-        raise ValueError(f"Unsupported file format: {ext}")
+
+def load_dataframe(filepath: str | Path,
+                   sheet_name: Optional[str | int] = None,
+                   sep: Optional[str] = None,
+                   encoding: Optional[str] = None,
+                   **_: Any) -> pd.DataFrame:
+    """
+    Universal table loader with safe fallbacks.
+
+    Supports:
+      .csv, .txt, .tsv
+      .xlsx, .xls          (needs openpyxl)
+      .parquet             (pyarrow or fastparquet)
+      .feather/.ft         (pyarrow)
+      .json
+      .pkl/.pickle
+      .sas7bdat/.xpt       (pyreadstat preferred; falls back to pandas.read_sas)
+      .sav                 (SPSS; pyreadstat)
+      .dta                 (Stata)
+    """
+    p = Path(filepath)
+    ext = p.suffix.lower()
+
+    # --- Delimited text ---
+    if ext in {".csv", ".txt"}:
+        # If sep is unspecified:
+        # - CSV: assume comma (keeps fast C engine, no warning)
+        # - TXT: keep sep=None (we'll use Python engine to infer)
+        eff_sep = sep if sep is not None else ("," if ext == ".csv" else None)
+        engine = "c" if eff_sep is not None else "python"
+        try:
+            return pd.read_csv(p, sep=eff_sep, encoding=encoding, engine=engine)
+        except UnicodeDecodeError:
+            return pd.read_csv(p, sep=eff_sep, encoding=encoding or "latin-1", engine=engine)
+
+    if ext == ".tsv":
+        try:
+            return pd.read_csv(p, sep="\t", encoding=encoding, engine="c")
+        except UnicodeDecodeError:
+            return pd.read_csv(p, sep="\t", encoding=encoding or "latin-1", engine="c")
+
+
+    # --- Excel ---
+    if ext in {".xlsx", ".xls"}:
+        try:
+            return pd.read_excel(p, sheet_name=0 if sheet_name is None else sheet_name)
+        except ImportError as e:
+            raise ModuleNotFoundError(
+                "openpyxl>=3.1 is required for Excel files. Install with: pip install openpyxl"
+            ) from e
+
+    # --- Columnar ---
+    if ext == ".parquet":
+        try:
+            return pd.read_parquet(p, engine="pyarrow")
+        except Exception:
+            return pd.read_parquet(p, engine="fastparquet")
+
+    if ext in {".feather", ".ft"}:
+        try:
+            import pyarrow  # noqa: F401
+        except ModuleNotFoundError as e:
+            raise ModuleNotFoundError(
+                "pyarrow is required for Feather files. Install with: pip install pyarrow"
+            ) from e
+        return pd.read_feather(p)
+
+    # --- Other common formats ---
+    if ext in {".pkl", ".pickle"}:
+        return pd.read_pickle(p)
+
+    if ext == ".json":
+        return pd.read_json(p, convert_dates=True)
+
+    # --- SAS / SPSS / Stata ---
+    if ext in {".sas7bdat", ".xpt"}:
+        try:
+            import pyreadstat  # type: ignore
+            if ext == ".sas7bdat":
+                df, _ = pyreadstat.read_sas7bdat(str(p))
+            else:
+                df, _ = pyreadstat.read_xport(str(p))
+            return df
+        except ModuleNotFoundError:
+            # best-effort fallback
+            return pd.read_sas(p)
+
+    if ext == ".sav":  # SPSS
+        try:
+            import pyreadstat  # type: ignore
+        except ModuleNotFoundError as e:
+            raise ModuleNotFoundError(
+                "pyreadstat is required for SPSS .sav files. Install with: pip install pyreadstat"
+            ) from e
+        df, _ = pyreadstat.read_sav(str(p))
+        return df
+
+    if ext == ".dta":  # Stata
+        return pd.read_stata(p)
+
+    raise ValueError(f"Unsupported file format: {ext} (path={p})")

tanml-0.1.7.dist-info/METADATA

@@ -0,0 +1,164 @@
+Metadata-Version: 2.4
+Name: tanml
+Version: 0.1.7
+Summary: Automated validation toolkit for tabular ML models—MRM, credit risk, insurance, and other regulated use cases.
+Author: Tanmay Sah, Dolly Sah
+Maintainer: Tanmay Sah, Dolly Sah
+License-Expression: MIT
+Project-URL: Homepage, https://github.com/tdlabs-ai/tanml
+Project-URL: Source, https://github.com/tdlabs-ai/tanml
+Project-URL: Issues, https://github.com/tdlabs-ai/tanml/issues
+Project-URL: Documentation, https://github.com/tdlabs-ai/tanml#readme
+Keywords: model validation,model risk management,model governance,SR 11-7,tabular ML,credit risk,insurance analytics,explainability,XAI,SHAP,stress testing,reporting,docx,streamlit,xgboost,lightgbm,catboost
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Classifier: Intended Audience :: Financial and Insurance Industry
+Classifier: Natural Language :: English
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3 :: Only
+Classifier: Programming Language :: Python :: 3.8
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Scientific/Engineering :: Information Analysis
+Classifier: Topic :: Scientific/Engineering :: Visualization
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=1.26
+Requires-Dist: scipy>=1.11
+Requires-Dist: pandas>=2.0
+Requires-Dist: scikit-learn>=1.3
+Requires-Dist: statsmodels>=0.14
+Requires-Dist: xgboost>=2.0
+Requires-Dist: lightgbm>=4.3
+Requires-Dist: catboost>=1.2
+Requires-Dist: shap>=0.44
+Requires-Dist: numba>=0.58
+Requires-Dist: matplotlib>=3.8
+Requires-Dist: seaborn>=0.13
+Requires-Dist: Pillow>=10.0
+Requires-Dist: python-docx>=1.1.2
+Requires-Dist: tzlocal>=5.0
+Requires-Dist: tqdm>=4.66
+Requires-Dist: pyarrow>=14.0
+Requires-Dist: openpyxl>=3.1
+Requires-Dist: pyreadstat>=1.2
+Requires-Dist: streamlit>=1.36
+Provides-Extra: dev
+Requires-Dist: pytest; extra == "dev"
+Requires-Dist: black; extra == "dev"
+Requires-Dist: isort; extra == "dev"
+Dynamic: license-file
+
+# TanML: Automated Model Validation Toolkit for Tabular Machine Learning
+
+[![Cite this repo](https://img.shields.io/badge/Cite-this_repo-blue)](https://github.com/tdlabs-ai/tanml#license--citation)
+[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)
+[![Downloads](https://pepy.tech/badge/tanml)](https://pepy.tech/project/tanml)
+
+**TanML** validates tabular ML models with a zero-config **Streamlit UI** and exports an audit-ready, **editable Word report (.docx)**. It covers data quality, correlation/VIF, performance, explainability (SHAP), and robustness/stress tests—built for regulated settings (MRM, credit risk, insurance, etc.).
+
+* **Status:** Beta (`0.x`)
+* **License:** MIT
+* **Python:** 3.8–3.12
+* **OS:** Linux / macOS / Windows (incl. WSL)
+
+---
+
+## Why TanML?
+
+* **Zero-config UI:** launch Streamlit, upload data, click **Run**—no YAML needed.
+* **Audit-ready outputs:** tables/plots + a polished DOCX your stakeholders can edit.
+* **Regulatory alignment:** supports common Model Risk Management themes (e.g., SR 11-7 style).
+* **Works with your stack:** scikit-learn, XGBoost/LightGBM/CatBoost, etc.
+
+---
+
+## Install
+
+```bash
+pip install tanml
+```
+
+## Quick Start (UI)
+
+```bash
+tanml ui
+```
+
+* Opens at **[http://127.0.0.1:8501](http://127.0.0.1:8501)**
+* **Upload limit ~1 GB** (preconfigured)
+* **Telemetry disabled by default**
+
+### In the app
+
+1. **Load data** — upload a cleaned CSV/XLSX/Parquet (optional: raw or separate Train/Test).
+2. **Select target & features** — target auto-suggested; features default to all non-target columns.
+3. **Pick a model** — choose library/algorithm (scikit-learn, XGBoost, LightGBM, CatBoost) and tweak params.
+4. **Run validation** — click **▶️ Refit & validate**.
+5. **Export** — click **⬇️ Download report** to get a **DOCX** (auto-selects classification/regression template).
+
+**Outputs**
+
+* Report: `./.ui_runs/<session>/tanml_report_*.docx`
+* Artifacts (CSV/PNGs): `./.ui_runs/<session>/artifacts/*`
+
+---
+
+## What TanML Checks
+
+* **Raw Data (optional):** rows/cols, missingness, duplicates, constant columns
+* **Data Quality & EDA:** summaries, distributions
+* **Correlation & Multicollinearity:** heatmap, top-pairs CSV, **VIF** table
+* **Performance**
+
+  * **Classification:** AUC, PR-AUC, KS, decile lift, confusion
+  * **Regression:** R², MAE, MSE/RMSE, error stats
+* **Explainability:** SHAP (auto explainer; configurable background size)
+* **Robustness/Stress Tests:** feature perturbations → delta-metrics
+* **Model Metadata:** model class, hyperparameters, features, training info
+
+---
+
+
+## Templates
+
+TanML ships DOCX templates (packaged in wheel & sdist):
+
+* `tanml/report/templates/report_template_cls.docx`
+* `tanml/report/templates/report_template_reg.docx`
+
+---
+
+
+## License & Citation
+
+**License:** MIT. See [LICENSE](https://github.com/tdlabs-ai/tanml/blob/main/LICENSE).  
+SPDX-License-Identifier: MIT
+
+© 2025 Tanmay Sah and Dolly Sah. You may use, modify, and distribute this software with appropriate attribution.
+
+### How to cite
+
+If TanML helps your work or publications, please cite:
+
+> Sah, T., & Sah, D. (2025). *TanML: Automated Model Validation Toolkit for Tabular Machine Learning* [Software]. Available at https://github.com/tdlabs-ai/tanml
+
+Or in BibTeX (version-agnostic):
+
+```bibtex
+@misc{tanml,
+  author = {Sah, Tanmay and Sah, Dolly},
+  title  = {TanML: Automated Model Validation Toolkit for Tabular Machine Learning},
+  year   = {2025},
+  note   = {Software; MIT License},
+  url    = {https://github.com/tdlabs-ai/tanml}
+}
+```
+
+A machine-readable citation file (`CITATION.cff`) is included for citation tools and GitHub’s “Cite this repository” button.

tanml-0.1.7.dist-info/RECORD

@@ -0,0 +1,54 @@
+tanml/__init__.py,sha256=GhTYFGQ77wbJM35lScp2XZKYnkNHTrdYozPoCZiCVSM,23
+tanml/check_runners/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+tanml/check_runners/base_runner.py,sha256=jibUcl6a7SQH9EKnPFZhU1t9FY-Bu_nkPjRfjW2T5x0,210
+tanml/check_runners/cleaning_repro_runner.py,sha256=fXlujWnuA0Hfzu9uQ31itU8eO-6wGD4eUImiR6kyHfg,817
+tanml/check_runners/correlation_runner.py,sha256=-TxktyjXi8IX4t0DCTe-9Ea5vVb-RcN6gBptbzUJdNQ,1975
+tanml/check_runners/data_quality_runner.py,sha256=IdhWYOtDBPkAwT2Aa6SYSiI2gkLEhkBXHAQop5ZFg0I,883
+tanml/check_runners/eda_runner.py,sha256=Gr5ZmgOvUilej-rhinsL9KmsIQOMWjd6jdOYUQTJ4os,623
+tanml/check_runners/explainability_runner.py,sha256=I40KVRoN1yEzZ6i1WZ6a1g89cPrRHpZ97lbfMEyOq2U,908
+tanml/check_runners/input_cluster_runner.py,sha256=0O8JJq3HOvNd3_nrZY8FevVYimI_Nl2Iftyyx56LVzc,1251
+tanml/check_runners/logistic_stats_runner.py,sha256=vTwnWHwUjl7ML2PKDteOgUc247D7xGXkKa8NfxUxYa4,7739
+tanml/check_runners/model_meta_runner.py,sha256=hOIEp6k98o0FwYjgmi_53Ni7H14jerKdf46wpxVkAkk,799
+tanml/check_runners/performance_runner.py,sha256=Q9r5mqj2ayOSfIN2SVaF4sm_Gc-pVyPfU3Q9NE9uZ8k,2567
+tanml/check_runners/raw_data_runner.py,sha256=qxX2Fst7u2VUwobtWr1StaDj8x5q1P01zCDhB0ewbdY,1531
+tanml/check_runners/regression_metrics_runner.py,sha256=wPDtGs713aX0rVRcUNsWVIybQee8MlHt6dxJxigRAFY,5774
+tanml/check_runners/rule_engine_runner.py,sha256=xgij_9S8kUSESgXeWGN7FnOD0hNmd3GAregxuD_x9nI,286
+tanml/check_runners/stress_test_runner.py,sha256=s4Ax7JnqGzqc3kw0wfCGt3LNIcZzZMX30Uk73zPDT34,1544
+tanml/check_runners/vif_runner.py,sha256=T7ho2jYnHKuz3P0R5CuoNB0LjPoH_Gs_FS0dcoyYbj4,2311
+tanml/checks/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+tanml/checks/base.py,sha256=g1GLXHWkE848bDKb4zXLHRqfgucVWFa-n6ZDT5vzs2I,570
+tanml/checks/cleaning_repro.py,sha256=LtPEkikjTY0NMsv1FZ8tOPUkaqF3yK_-DQdLCltMsIA,1795
+tanml/checks/correlation.py,sha256=IN86Yc-yUlBKNSaK-YTlGM4V2DH6BPZw5KHJVN-hE2c,10350
+tanml/checks/data_quality.py,sha256=bNrM469c_G-TxU-6ne2UYv5gwoUoyrab4lYbI16l7VU,981
+tanml/checks/eda.py,sha256=7fOIBhCg0oiK4SL22WcKHK8vL3s-YwypiAnZDwoq3lo,2515
+tanml/checks/input_cluster.py,sha256=93RvE-vd_elt4xqnkCkHmHH6NWPOtJF7ylb5mPTsagA,3990
+tanml/checks/logit_stats.py,sha256=OnL2_M4bg-T5RE5Rv_wMJNkaX4Ox1ihSk428_7OJ15w,6958
+tanml/checks/model_contents.py,sha256=jcS4GrUoaQVUBf9pFwvtv9KKvDO5vZroiPz8mOmyAGo,1263
+tanml/checks/model_meta.py,sha256=x2RnT3kUWY4ERWfM9cItpaIJX8Fca9IQByR9zPN2V20,1881
+tanml/checks/performance.py,sha256=vQ_FeIcr3ASDhVd97y2JKMEMfJUrfCRnUBZh1ZI6rCs,3400
+tanml/checks/performance_classification.py,sha256=2MnMeQEBVswCOjT7hFK7gjRBiXmhLgriA6_qw_9jsXk,11982
+tanml/checks/raw_data.py,sha256=asjyURkvBQ1qOfTJkIZblcyELf1HO044jRo1BZGsPOQ,2862
+tanml/checks/regression_metrics.py,sha256=736Hk6tmCwR8uFUJF1Z5kszFiT_qHM86jBPGveb_BkY,6158
+tanml/checks/rule_engine.py,sha256=bqGd0-2MlSPGGUFPyMxgHiwVTvw1Iu7NFqUzl_-0iIg,1694
+tanml/checks/stress_test.py,sha256=64fVn96uzGg7_qvPMwiWhYRSr4FflFMkkjXQXmv2YB0,7460
+tanml/checks/vif.py,sha256=0tghDaiG8z4frgP5OHj3ctgsH4Mak1N7Docu6OcG8JA,1902
+tanml/checks/explainability/shap_check.py,sha256=Di5WD3jUgs1TEwnhCoEU0dhvMlBUf0I3tYNpsjST1wk,12829
+tanml/cli/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+tanml/cli/main.py,sha256=A9W7RXYbnvw_O-jJDbwvkAOL4NQF93XNsn3PlzNuqZg,3929
+tanml/config_templates/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+tanml/engine/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+tanml/engine/check_agent_registry.py,sha256=CrRAL_BnsXcivXLQydka2R8emHrkHOqZDSzuqZNmknc,2419
+tanml/engine/core_engine_agent.py,sha256=-DE9ehdY60W2IAzNMnv-BTOf3KPJobX8p5dBcWH1aiE,9334
+tanml/models/registry.py,sha256=SMAwKTLVIFYQ_KrYsaxm-dmSfyE_zk1Mo1AL8-JZt1w,15467
+tanml/report/report_builder.py,sha256=_UuaLxJOuvfXZjOWIrI478G8jMUy2CxQNi2YDR-8FYI,55915
+tanml/report/templates/report_template_cls.docx,sha256=1pAKIvC8nG2JGN3H7SpU10QVaeDNgJI1aOi1SH6TCzs,30199
+tanml/report/templates/report_template_reg.docx,sha256=bjr-18WFD-KwtRtNcbRipstyNa29z7hF-JHp86OMKVg,29673
+tanml/ui/app.py,sha256=n6seLRuoWSF6sK6arewyhhEyfEhNlGeyhourRAB_loc,47682
+tanml/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+tanml/utils/data_loader.py,sha256=_Ja3XSONwL_NVwpxuBjLacXWX5yPTN9H0moFxSVqGlw,3713
+tanml-0.1.7.dist-info/licenses/LICENSE,sha256=e6xQyG7SdWiD4cLlj7rFdMrjG6H7ABglGOrooZxWLKQ,1102
+tanml-0.1.7.dist-info/METADATA,sha256=gy0-NjrrTIIl9J0kFL2iNTI07AYuNSa-GEf_Iq1wSHo,6118
+tanml-0.1.7.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+tanml-0.1.7.dist-info/entry_points.txt,sha256=WUM_y0uRIL7iXPcxK69Bn5mKZXnDshWDbLyohjND1IE,46
+tanml-0.1.7.dist-info/top_level.txt,sha256=81dIhCm6opwY6E7Pb9G1kdIVmYrUkXX4PaYhQ873gIE,6
+tanml-0.1.7.dist-info/RECORD,,

tanml/cli/arg_parser.py

@@ -1,31 +0,0 @@
-import argparse
-
-def parse_args():
-    parser = argparse.ArgumentParser(description="Run TanML model validation toolkit")
-    subparsers = parser.add_subparsers(dest="command", required=True)
-
-
-    validate_parser = subparsers.add_parser("validate", help="Run validation checks and generate report")
-    validate_parser.add_argument("--model", required=False,
-                        help="Model path: .pkl for sklearn/xgb, .csv for SAS or R logistic")
-    validate_parser.add_argument("--raw", required=False, help="Path to raw input data file")
-    validate_parser.add_argument("--cleaned", required=False, help="Path to cleaned input data file")
-    validate_parser.add_argument("--rules", required=True, help="Path to rules.yaml config file")
-    validate_parser.add_argument("--target", required=False, help="Target column name (optional)")
-    validate_parser.add_argument("--features", required=False, help="Comma-separated list of features")
-    validate_parser.add_argument(
-        "--report_path",
-        type=str,
-        default="reports/final_report.docx",
-        help="Path to output DOCX report. Example: --report_path my_reports/run1.docx"
-    )
-
-
-    init_parser = subparsers.add_parser("init", help="Generate starter rules.yaml file")
-    init_parser.add_argument("--scenario", required=True, choices=["A", "B", "C"],
-                             help="Choose validation scenario: A (single model), B (multiple segments), C (single dataset + segment column)")
-    init_parser.add_argument(
-        "--output", type=str, default="rules.yaml",
-        help="Destination path for rules YAML file (default: rules.yaml)"
-    )
-    return parser.parse_args()

tanml/cli/init_cmd.py

@@ -1,8 +0,0 @@
-from tanml.utils.yaml_generator import generate_rules_yaml
-
-def run_init(scenario, dest_path="rules.yaml", overwrite=False):
-    try:
-        generate_rules_yaml(scenario=scenario, dest_path=dest_path, overwrite=overwrite)
-
-    except Exception as e:
-        print(f"❌ Failed to create YAML: {e}")

tanml/cli/validate_cmd.py

@@ -1,7 +0,0 @@
-# tanml/cli/validate_cmd.py
-
-from tanml.validate import validate_from_yaml
-
-def run_validate(rules_path):
-    print(f"🧪 Starting validation using rules from: {rules_path}")
-    validate_from_yaml(rules_path)

tanml/config_templates/rules_multiple_models_datasets.yaml

@@ -1,144 +0,0 @@
-# ============================================================
-# TanML Validation Configuration File: Scenario B
-# ------------------------------------------------------------
-# 🧪 Scenario: One model and one cleaned dataset per segment
-#
-# ✅ Required:
-#   - Define segment-wise cleaned data and model path (or retrain)
-#   - Choose ONE model source strategy (Option A or Option B)
-#   - Set input features and target column
-#   - Optional: provide global raw data
-#   - Adjust thresholds and check options as needed
-# ============================================================
-
-# ------------------------------------------
-# REQUIRED: Model Input Schema
-# ------------------------------------------
-model:
-  features:
-    - feature_0    # 👉 replace with actual feature names used in all segment models
-    - feature_1    # 👉 remove or add more lines if needed
-    - feature_2
-  target: default_flag   # 👉 replace with your actual target column
-
-# ------------------------------------------
-# OPTIONAL: Raw Data Path
-# ------------------------------------------
-paths:
-  raw_data: data/raw.csv                # 👉 optional – provide full path if you have original raw dataset otherwise use null
-
-output:
-  report_path_template: /absolute/path/to/output/{segment} # 👉 Output path template – {segment} will be replaced dynamically, it is folder path
-
-# ------------------------------------------
-# ✅ OPTION A — Pretrained Model per Segment (.pkl)
-# ------------------------------------------
-# Use this option when each segment has its own trained model and cleaned dataset.
-# You must provide:
-#   - model: path to the pretrained `.pkl` file
-#   - cleaned: path to the cleaned CSV used for that segment
-#
-# 👉 Comment out the OPTION B block below if using this
-# ------------------------------------------
-segment:
-  runs:
-    segment_A:                           # 👉 rename (e.g., high_risk, bronze, tier_1)
-      model: models/logistic/model_b_segment1.pkl
-      cleaned: data/scenario_b_segment1.csv
-
-    segment_B:
-      model: models/logistic/model_b_segment1.pkl
-      cleaned: data/scenario_b_segment2.csv
-
-# ------------------------------------------
-# 🔁 OPTION B — Retrain Model from Cleaned Data (Per Segment)
-# ------------------------------------------
-# Use this when you want to retrain a model for each segment
-# using the corresponding cleaned dataset and shared model config.
-#
-# 👉 If using this, comment out OPTION A (i.e., segment.runs[].model lines)
-# 👉 Each segment must still have its own cleaned dataset
-# ------------------------------------------
-
-# segment:
-#   runs:
-#     segment_A:
-#       cleaned: data/cleaned_a.csv
-#     segment_B:
-#       cleaned: data/cleaned_b.csv
-
-# ------------------------------------------
-# MODEL SOURCE CONFIG
-# ------------------------------------------
-# This tells TanML how to load or build the model(s) for each segment.
-#
-# 👉 If using pretrained models (Option A): set `from_pickle: true`
-# 👉 If retraining per segment (Option B): set `from_pickle: false`
-# ------------------------------------------
-model_source:
-  from_pickle: true     # 👉 change to false if using retraining (Option B)
-  type: LogisticRegression
-  module: sklearn.linear_model
-  hyperparameters:
-    penalty: "l2"
-    solver: "liblinear"
-    random_state: 42
-    class_weight: "balanced"
-    max_iter: 100
-
-# ------------------------------------------
-# PERFORMANCE THRESHOLDS
-# ------------------------------------------
-auc_roc:
-  min: 0.60
-
-f1:
-  min: 0.60
-
-ks:
-  min: 0.20
-
-# ------------------------------------------
-# VALIDATION CHECKS
-# ------------------------------------------
-
-EDACheck:
-  enabled: true
-  max_plots: -1             # -1 = all numeric; or set number of columns
-
-correlation:
-  enabled: true
-
-VIFCheck:
-  enabled: true
-  
-raw_data_check:
-  enabled: true
-
-model_meta:
-  enabled: true
-
-# ------------------------------------------
-# STRESS TESTING (Robustness Check)
-# ------------------------------------------
-StressTestCheck:
-  enabled: true
-  epsilon: 0.01             # ➜ 1% noise
-  perturb_fraction: 0.2     # ➜ 20% of rows
-
-# ------------------------------------------
-# INPUT CLUSTER COVERAGE
-# ------------------------------------------
-InputClusterCoverageCheck:
-  enabled: true
-  n_clusters: 5             # ➜ fixed clusters for coverage bar chart
-  max_k: 10                 # ➜ elbow method search (if needed)
-
-# ------------------------------------------
-# EXPLAINABILITY
-# ------------------------------------------
-explainability:
-  shap:
-    enabled: true
-    background_sample_size: 100   # ➜ SHAP explainer training background
-    test_sample_size: 200         # ➜ test rows to explain

tanml/config_templates/rules_one_dataset_segment_column.yaml

@@ -1,140 +0,0 @@
-# ============================================================
-# TanML Validation Configuration File: Scenario C
-# ------------------------------------------------------------
-# 🧪 Scenario: One dataset with a segment column, one model per segment
-#
-# ✅ Required:
-#   - Provide cleaned data (includes all segments)
-#   - Define the segment column used to split the data
-#   - Choose ONE model source strategy (Option A or Option B)
-#   - Set input features and target column
-#   - Optional: provide global raw data
-#   - Adjust thresholds and check options as needed
-# ============================================================
-
-# ------------------------------------------
-# REQUIRED: Model Input Schema
-# ------------------------------------------
-model:
-  features:
-    - feature_0    # 👉 replace with actual feature names used across all segment models
-    - feature_1
-    - feature_2
-  target: default_flag   # 👉 replace with your actual target column
-
-# ------------------------------------------
-# REQUIRED: File Paths
-# ------------------------------------------
-paths:
-  cleaned_data: data/cleaned.csv          # 👉 path to full cleaned dataset (includes all segments)
-  raw_data: data/raw.csv                  # 👉 optional — use null if raw data not available
-
-# ------------------------------------------
-# OUTPUT CONFIGURATION
-# ------------------------------------------
-# Path template where validation reports will be saved.
-# 👉 Use `{segment}` as a placeholder for the segment name.
-# ------------------------------------------
-output:
-  report_path_template: reports/scenario_c/{segment}_report.docx  # 👉 customize this path as needed
-
-# ------------------------------------------
-# ✅ OPTION A — Pretrained Models per Segment (.pkl)
-# ------------------------------------------
-# Provide one model per segment (already trained)
-# This is the default option.
-# 👉 Comment out the OPTION B block below if using this
-# ------------------------------------------
-segment:
-  column: customer_segment                # 👉 column used to split segments
-
-  runs:
-    segment_A:
-      model: models/logistic/model_a.pkl
-    segment_B:
-      model: models/logistic/model_b.pkl
-
-# ------------------------------------------
-# 🔁 OPTION B — Retrain Models per Segment from Cleaned Data
-# ------------------------------------------
-# Use this if you want TanML to retrain a model for each segment
-# from the common cleaned dataset.
-# 👉 Comment out the OPTION A block above if using this
-# 👉 `segment.runs` must list segment values (no model paths needed)
-# ------------------------------------------
-# segment:
-#   column: customer_segment
-#   runs:
-#     segment_A: {}
-#     segment_B: {}
-
-# ------------------------------------------
-# MODEL SOURCE CONFIGURATION
-# ------------------------------------------
-# 👉 If using pretrained models: set `from_pickle: true`
-# 👉 If retraining per segment: set `from_pickle: false`
-# ------------------------------------------
-model_source:
-  from_pickle: true
-  type: LogisticRegression
-  module: sklearn.linear_model
-  hyperparameters:
-    penalty: "l2"
-    solver: "liblinear"
-    random_state: 42
-    class_weight: "balanced"
-    max_iter: 100
-
-# ------------------------------------------
-# PERFORMANCE THRESHOLDS
-# ------------------------------------------
-auc_roc:
-  min: 0.60
-f1:
-  min: 0.60
-ks:
-  min: 0.20
-
-# ------------------------------------------
-# VALIDATION CHECKS
-# ------------------------------------------
-EDACheck:
-  enabled: true
-  max_plots: -1
-
-correlation:
-  enabled: true
-
-VIFCheck:
-  enabled: true
-
-raw_data_check:
-  enabled: true
-
-model_meta:
-  enabled: true
-
-# ------------------------------------------
-# STRESS TESTING (Robustness Check)
-# ------------------------------------------
-StressTestCheck:
-  enabled: true
-  epsilon: 0.01
-  perturb_fraction: 0.2
-
-# ------------------------------------------
-# INPUT CLUSTER COVERAGE
-# ------------------------------------------
-InputClusterCoverageCheck:
-  enabled: true
-  n_clusters: 5
-  max_k: 10
-
-# ------------------------------------------
-# EXPLAINABILITY
-# ------------------------------------------
-explainability:
-  shap:
-    enabled: true
-    background_sample_size: 100
-    test_sample_size: 200