Marimo Notebook update

Author: hnlearndev

.gitignore

@@ -65,6 +65,7 @@ notebooks/outputs/
 notebooks/*.parquet
 notebooks/*.csv
 notebooks/archived
+notebooks/__marimo__
 
 # Ruff cache
 .notebooks/.ruff_cache/

README.md

@@ -5,6 +5,7 @@
 **Some notes**:
 
 - The main reference is [Basic term S](https://lifelib.io/libraries/basiclife/BasicTerm_S.html). Model S, M and SC are basically the same.
+- The computation heavy work load is performed with Rust.
 - Post run analysis is stored at [notebooks](notebooks) folder. This is a [Marimo Notebook](https://marimo.io/) notebook. Its purpose is to explain a practical smoothing technique after crude premium calculation is performed by Rust model. The method is  reproducible and  reduces labor intensity from manual practise.
 
 **What can be taken out from this repo?**

notebooks/analysis.py

@@ -43,11 +43,16 @@ def _(mo):
 
     This notebook shows how the premium rates are smoothed after being calculated using Rust model.
 
-    The intension at first is to used Rust for analysis. However, at the time of this is produced, due to its Rust language's nature (compiled), it takes significant time to produce the output. On another hand, Python reactive notebook is more advanced and packed with featured. Hence, Python is used for post anlysis.
+    The intension at first is to used Rust for analysis.
+    However, at the time of production, Rust language compilation nature takes significant time to produce the output in reactive notebook environment.
+    On another hand, Python reactive notebook is more advanced and packed with featured.
+    Hence, Python is used for post analysis.
 
-    After premium rate is calculated using Rust model, the premium rates are still jumpy at several model points. This notebook shows how the premium rate is smoothed using common p-sline technique.
+    What is done in this notebook?
+    After premium rate is calculated using Rust model, the premium rates are still jumpy at several model points.
+    This notebook shows how the premium rate is smoothed using common p-sline technique.
 
-    In practise, it is observed that many actuaries are performing this step manually with interpolation. However, it is not recommended as not being reproducible and labor intensive.
+    In practise, this step is usually performed manually with interpolation. However, it is not recommended as not being reproducible and labor-intensive.
 
     ## Crude premium rate
 
@@ -119,7 +124,7 @@ def _(Ridge, SplineTransformer, make_pipeline, np, plt, prem_rate_pd):
     smoothed_points = []
 
     for term, group in prem_rate_pd.groupby("term"):
-        # 
+        #
         group = group.sort_values("age")
         X = group["age"].values.reshape(-1, 1)
         y = group["ave_prem_rate"].values
@@ -131,10 +136,10 @@ def _(Ridge, SplineTransformer, make_pipeline, np, plt, prem_rate_pd):
             Ridge(alpha=0.0)
         )
         model.fit(X, y)
-    
+
         # Predict at each observed age
         y_smooth = model.predict(X)
-    
+
         # Store smoothed values
         group_smoothed = group.copy()
         group_smoothed["smoothed_prem_rate"] = y_smooth
@@ -203,7 +208,7 @@ def _(pl, smoothed_pl):
 def _(mo):
     mo.md(
         r"""
-    As observed there is still one points that is has a negative slope. We can either iterate over different subset of variable to smooth out the curve. 
+    As observed there is still one points that is has a negative slope. We can either iterate over different subset of variable to smooth out the curve.
 
     However, I make an objective decision if tolerance is small enough, it will be ignored. Else, we will obtain the rate from the previous period to make the difference 0.
     """

src/main.rs

@@ -48,7 +48,7 @@ fn main() -> PolarsResult<()> {
 
     let multi_run_results = run_setups.projection_runs()?;
 
-    multi_run_results.export("src/results/third_test")?;
+    multi_run_results.export("src/results/first_test")?;
 
     // Print the time taken for the runs
     let duration = start.elapsed();