kaushik7425/Cable_Optimization

1import streamlit as st
2import pandas as pd
3import seaborn as sns
4import matplotlib.pyplot as plt
5import numpy as np
6
7from sklearn.linear_model import ElasticNet, Ridge, LinearRegression
8from sklearn.model_selection import train_test_split, GridSearchCV
9from sklearn.metrics import r2_score
10
11st.set_page_config(layout="wide")
12
13# ================= LOAD DATA =================
14@st.cache_data
15def load_data():
16    df = pd.read_csv("cabeldata.csv")
17    df.columns = [c.strip() for c in df.columns]
18    return df
19
20df = load_data()
21target = "Simulated Capacitance(pF) for 4.2ft"
22
23numeric_cols = df.select_dtypes(include="number").columns.tolist()
24numeric_cols.remove(target)
25
26# ================= SIDEBAR =================
27st.sidebar.title("Model & Physics Control")
28
29model_choice = st.sidebar.selectbox(
30    "Select Model",
31    ["ElasticNet (Recommended)", "Ridge", "Linear Regression"]
32)
33
34selected_features = st.sidebar.multiselect(
35    "Select Input Physics",
36    numeric_cols,
37    default=[
38        c for c in [
39            "Conductor OD(mm)",
40            "Conductor Dielectric OD(mm)",
41            "insulated diameter mil",
42            "Packing Ratio %",
43            "TPI",
44            "OD conductor bundle",
45            "Desire shield OD",
46            "Dielectric material PN 155C/ .7 loss tangent",
47            "Free air capacitance for 4.2ft single Twisted Pair",
48            "Free air capacitance for single Twisted Pair pf/ft"
49        ] if c in numeric_cols
50    ]
51)
52
53remove_outliers = st.sidebar.checkbox("Remove Outliers", True)
54train_btn = st.sidebar.button("Train Model")
55
56# ================= TRAIN =================
57if train_btn:
58
59    data = df[selected_features + [target]].dropna()
60    X = data[selected_features]
61    y = data[target]
62
63    if remove_outliers:
64        base = LinearRegression()
65        base.fit(X, y)
66        residuals = y - base.predict(X)
67        mask = residuals.abs() < 10
68        X = X[mask]
69        y = y[mask]
70
71    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=42)
72
73    if model_choice == "Linear Regression":
74        model = LinearRegression()
75        model.fit(X_train, y_train)
76
77    elif model_choice == "Ridge":
78        grid = GridSearchCV(Ridge(), {"alpha":[0.01,0.1,1,10,50]}, cv=5, scoring="r2")
79        grid.fit(X_train, y_train)
80        model = grid.best_estimator_
81        params = grid.best_params_
82
83    else:
84        grid = GridSearchCV(
85            ElasticNet(max_iter=10000),
86            {"alpha":[0.001,0.01,0.1,1,10],"l1_ratio":[0.1,0.3,0.5,0.7,0.9]},
87            cv=5, scoring="r2"
88        )
89        grid.fit(X_train, y_train)
90        model = grid.best_estimator_
91        params = grid.best_params_
92
93    y_pred = model.predict(X_test)
94
95    st.session_state.model = model
96    st.session_state.features = selected_features
97    st.session_state.y_test = y_test
98    st.session_state.y_pred = y_pred
99    st.session_state.r2 = r2_score(y_test, y_pred)
100    st.session_state.model_choice = model_choice
101    st.session_state.params = params if "params" in locals() else None
102
103# ================= UI =================
104st.title("RF Cable Capacitance Digital Twin")
105
106if "model" in st.session_state:
107    st.success(f"{st.session_state.model_choice} | Test R² = {st.session_state.r2}")
108    if st.session_state.params:
109        st.info(f"Model parameters: {st.session_state.params}")
110else:
111    st.warning("Select physics + model and click Train")
112
113# ================= PREDICT =================
114if "model" in st.session_state:
115    st.subheader("Predict Cable")
116
117    inputs = {}
118    cols = st.columns(3)
119    for i,f in enumerate(st.session_state.features):
120        inputs[f] = cols[i%3].number_input(f, value=0.0, step=0.0000001, format="%.10f")
121
122    if st.button("Predict Capacitance"):
123        row = pd.DataFrame([inputs])
124        pred = st.session_state.model.predict(row)[0]
125        st.success(f"Predicted Capacitance = {pred} pF")
126
127# ================= PLOTS =================
128if "model" in st.session_state:
129    st.subheader("Diagnostics")
130
131    plot = st.selectbox("Select Plot",
132        ["None","Correlation Heatmap","Predicted vs Actual","Residuals","Feature vs Capacitance","GGPlot Smooth"]
133    )
134
135    if plot=="Correlation Heatmap":
136        fig,ax=plt.subplots(figsize=(10,6))
137        sns.heatmap(df[st.session_state.features+[target]].corr(), cmap="coolwarm", center=0, ax=ax)
138        st.pyplot(fig)
139
140    elif plot=="Predicted vs Actual":
141        fig,ax=plt.subplots()
142        ax.scatter(st.session_state.y_test, st.session_state.y_pred)
143        ax.plot([st.session_state.y_test.min(),st.session_state.y_test.max()],
144                [st.session_state.y_test.min(),st.session_state.y_test.max()])
145        st.pyplot(fig)
146
147    elif plot=="Residuals":
148        fig,ax=plt.subplots()
149        ax.scatter(st.session_state.y_test, st.session_state.y_test-st.session_state.y_pred)
150        ax.axhline(0)
151        st.pyplot(fig)
152
153    elif plot=="Feature vs Capacitance":
154        f=st.selectbox("Select Feature",st.session_state.features)
155        fig,ax=plt.subplots()
156        ax.scatter(df[f],df[target])
157        st.pyplot(fig)
158
159    elif plot=="GGPlot Smooth":
160        f=st.selectbox("Select Feature",st.session_state.features)
161        x=df[f]; y=df[target]
162        z=np.polyfit(x,y,3)
163        xp=np.linspace(x.min(),x.max(),200)
164        yp=np.polyval(z,xp)
165        fig,ax=plt.subplots()
166        ax.scatter(x,y,alpha=0.4)
167        ax.plot(xp,yp,color="red")
168        st.pyplot(fig)
169