shahin/EDSS-calc
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

optimize with new column names

Browse Source
This commit is contained in:
Shahin Ramezanzadeh
2026-01-19 02:29:38 +01:00
parent a1a8abfb8e
commit ce3baff6cc
1 changed files with 49 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

98
Data/show_plots.py
View File

@@ -93,19 +93,19 @@ from sklearn.metrics import confusion_matrix, classification_report
import seaborn as sns
# Load your data from TSV file
file_path = '/home/shahin/Lab/Doktorarbeit/Barcelona/Data/join_results_unique.tsv'
file_path = '/home/shahin/Lab/Doktorarbeit/Barcelona/Data/Join_edssandsub.tsv'
df = pd.read_csv(file_path, sep='\t')
# Replace comma with dot for numeric conversion in GT_EDSS and LLM_Results
df['GT_EDSS'] = df['GT_EDSS'].astype(str).str.replace(',', '.')
df['LLM_Results'] = df['LLM_Results'].astype(str).str.replace(',', '.')
# Replace comma with dot for numeric conversion in GT.EDSS and result.EDSS
df['GT.EDSS'] = df['GT.EDSS'].astype(str).str.replace(',', '.')
df['result.EDSS'] = df['result.EDSS'].astype(str).str.replace(',', '.')
# Convert to float (handle invalid entries gracefully)
df['GT_EDSS'] = pd.to_numeric(df['GT_EDSS'], errors='coerce')
df['LLM_Results'] = pd.to_numeric(df['LLM_Results'], errors='coerce')
df['GT.EDSS'] = pd.to_numeric(df['GT.EDSS'], errors='coerce')
df['result.EDSS'] = pd.to_numeric(df['result.EDSS'], errors='coerce')
# Drop rows where either column is NaN
df_clean = df.dropna(subset=['GT_EDSS', 'LLM_Results'])
df_clean = df.dropna(subset=['GT.EDSS', 'result.EDSS'])
# For confusion matrix, we need to categorize the values
# Let's create categories up to 10 (0-1, 1-2, 2-3, ..., 9-10)
@@ -136,14 +136,14 @@ def categorize_edss(value):
return '10+'
# Create categorical versions
df_clean['GT_EDSS_cat'] = df_clean['GT_EDSS'].apply(categorize_edss)
df_clean['LLM_Results_cat'] = df_clean['LLM_Results'].apply(categorize_edss)
df_clean['GT.EDSS_cat'] = df_clean['GT.EDSS'].apply(categorize_edss)
df_clean['result.EDSS_cat'] = df_clean['result.EDSS'].apply(categorize_edss)
# Remove any NaN categories
df_clean = df_clean.dropna(subset=['GT_EDSS_cat', 'LLM_Results_cat'])
df_clean = df_clean.dropna(subset=['GT.EDSS_cat', 'result.EDSS_cat'])
# Create confusion matrix
cm = confusion_matrix(df_clean['GT_EDSS_cat'], df_clean['LLM_Results_cat'],
cm = confusion_matrix(df_clean['GT.EDSS_cat'], df_clean['result.EDSS_cat'],
labels=['0-1', '1-2', '2-3', '3-4', '4-5', '5-6', '6-7', '7-8', '8-9', '9-10'])
# Plot confusion matrix
@@ -151,15 +151,15 @@ plt.figure(figsize=(10, 8))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
xticklabels=['0-1', '1-2', '2-3', '3-4', '4-5', '5-6', '6-7', '7-8', '8-9', '9-10'],
yticklabels=['0-1', '1-2', '2-3', '3-4', '4-5', '5-6', '6-7', '7-8', '8-9', '9-10'])
plt.title('Confusion Matrix: GT_EDSS vs interferred EDSS (Categorized 0-10)')
plt.xlabel('LLM_Results Category')
plt.ylabel('GT_EDSS Category')
plt.title('Confusion Matrix: Ground truth EDSS vs interferred EDSS (Categorized 0-10)')
plt.xlabel('LLM Generated EDSS')
plt.ylabel('Ground Truth EDSS')
plt.tight_layout()
plt.show()
# Print classification report
print("Classification Report:")
print(classification_report(df_clean['GT_EDSS_cat'], df_clean['LLM_Results_cat']))
print(classification_report(df_clean['GT.EDSS_cat'], df_clean['result.EDSS_cat']))
# Print raw counts
print("\nConfusion Matrix (Raw Counts):")
@@ -177,7 +177,7 @@ from sklearn.metrics import confusion_matrix
import numpy as np
# Load your data from TSV file
file_path ='/home/shahin/Lab/Doktorarbeit/Barcelona/Data/join_results_unique.tsv'
file_path ='/home/shahin/Lab/Doktorarbeit/Barcelona/Data/Join_edssandsub.tsv'
df = pd.read_csv(file_path, sep='\t')
@@ -209,29 +209,29 @@ def safe_bool_convert(series):
return converted
# Convert columns safely
if 'LLM_klassifizierbar' in df.columns:
print("\nLLM_klassifizierbar column info:")
print(df['LLM_klassifizierbar'].head(10))
print("Unique values:", df['LLM_klassifizierbar'].unique())
if 'result.klassifizierbar' in df.columns:
print("\nresult.klassifizierbar column info:")
print(df['result.klassifizierbar'].head(10))
print("Unique values:", df['result.klassifizierbar'].unique())
df['LLM_klassifizierbar'] = safe_bool_convert(df['LLM_klassifizierbar'])
df['result.klassifizierbar'] = safe_bool_convert(df['result.klassifizierbar'])
print("After conversion:")
print(df['LLM_klassifizierbar'].value_counts())
print(df['result.klassifizierbar'].value_counts())
if 'GT_klassifizierbar' in df.columns:
print("\nGT_klassifizierbar column info:")
print(df['GT_klassifizierbar'].head(10))
print("Unique values:", df['GT_klassifizierbar'].unique())
if 'GT.klassifizierbar' in df.columns:
print("\nGT.klassifizierbar column info:")
print(df['GT.klassifizierbar'].head(10))
print("Unique values:", df['GT.klassifizierbar'].unique())
df['GT_klassifizierbar'] = safe_bool_convert(df['GT_klassifizierbar'])
df['GT.klassifizierbar'] = safe_bool_convert(df['GT.klassifizierbar'])
print("After conversion:")
print(df['GT_klassifizierbar'].value_counts())
print(df['GT.klassifizierbar'].value_counts())
# Create bar chart showing only True values for klassifizierbar
if 'LLM_klassifizierbar' in df.columns and 'GT_klassifizierbar' in df.columns:
if 'result.klassifizierbar' in df.columns and 'GT.klassifizierbar' in df.columns:
# Get counts for True values only
llm_true_count = df['LLM_klassifizierbar'].sum()
gt_true_count = df['GT_klassifizierbar'].sum()
llm_true_count = df['result.klassifizierbar'].sum()
gt_true_count = df['GT.klassifizierbar'].sum()
# Plot using matplotlib directly
fig, ax = plt.subplots(figsize=(8, 6))
@@ -266,11 +266,11 @@ if 'LLM_klassifizierbar' in df.columns and 'GT_klassifizierbar' in df.columns:
plt.show()
# Create confusion matrix if both columns exist
if 'LLM_klassifizierbar' in df.columns and 'GT_klassifizierbar' in df.columns:
if 'result.klassifizierbar' in df.columns and 'GT.klassifizierbar' in df.columns:
try:
# Ensure both columns are boolean
llm_bool = df['LLM_klassifizierbar'].fillna(False).astype(bool)
gt_bool = df['GT_klassifizierbar'].fillna(False).astype(bool)
llm_bool = df['result.klassifizierbar'].fillna(False).astype(bool)
gt_bool = df['GT.klassifizierbar'].fillna(False).astype(bool)
cm = confusion_matrix(gt_bool, llm_bool)
@@ -295,7 +295,7 @@ if 'LLM_klassifizierbar' in df.columns and 'GT_klassifizierbar' in df.columns:
# Show final data info
print("\nFinal DataFrame info:")
print(df[['LLM_klassifizierbar', 'GT_klassifizierbar']].info())
print(df[['result.klassifizierbar', 'GT.klassifizierbar']].info())
##
@@ -312,23 +312,23 @@ import numpy as np
file_path = '/home/shahin/Lab/Doktorarbeit/Barcelona/Data/join_results_unique.tsv'
df = pd.read_csv(file_path, sep='\t')
# Replace comma with dot for numeric conversion in GT_EDSS and LLM_Results
df['GT_EDSS'] = df['GT_EDSS'].astype(str).str.replace(',', '.')
df['LLM_Results'] = df['LLM_Results'].astype(str).str.replace(',', '.')
# Replace comma with dot for numeric conversion in GT.EDSS and result.EDSS
df['GT.EDSS'] = df['GT.EDSS'].astype(str).str.replace(',', '.')
df['result.EDSS'] = df['result.EDSS'].astype(str).str.replace(',', '.')
# Convert to float (handle invalid entries gracefully)
df['GT_EDSS'] = pd.to_numeric(df['GT_EDSS'], errors='coerce')
df['LLM_Results'] = pd.to_numeric(df['LLM_Results'], errors='coerce')
df['GT.EDSS'] = pd.to_numeric(df['GT.EDSS'], errors='coerce')
df['result.EDSS'] = pd.to_numeric(df['result.EDSS'], errors='coerce')
# Drop rows where either column is NaN
df_clean = df.dropna(subset=['GT_EDSS', 'LLM_Results'])
df_clean = df.dropna(subset=['GT.EDSS', 'result.EDSS'])
# 1. DEFINE CATEGORY ORDER
# This ensures the X-axis is numerically logical (0-1 comes before 1-2)
category_order = ['0-1', '1-2', '2-3', '3-4', '4-5', '5-6', '6-7', '7-8', '8-9', '9-10', '10+']
# Convert the column to a Categorical type with the specific order
df_clean['GT_EDSS_cat'] = pd.Categorical(df_clean['GT_EDSS'].apply(categorize_edss),
df_clean['GT.EDSS_cat'] = pd.Categorical(df_clean['GT.EDSS'].apply(categorize_edss),
categories=category_order,
ordered=True)
@@ -338,16 +338,16 @@ plt.figure(figsize=(14, 8))
# Assigning x to 'hue' allows Seaborn to generate a legend automatically
box_plot = sns.boxplot(
data=df_clean,
x='GT_EDSS_cat',
y='LLM_Results',
hue='GT_EDSS_cat', # Added hue
x='GT.EDSS_cat',
y='result.EDSS',
hue='GT.EDSS_cat', # Added hue
palette='viridis',
linewidth=1.5,
legend=True # Ensure legend is enabled
)
# 3. CUSTOMIZE PLOT
plt.title('Distribution of LLM_Results by GT_EDSS Category', fontsize=18, pad=20)
plt.title('Distribution of result.EDSS by GT.EDSS Category', fontsize=18, pad=20)
plt.xlabel('Ground Truth EDSS Category', fontsize=14)
plt.ylabel('LLM Predicted EDSS', fontsize=14)
@@ -545,7 +545,7 @@ def create_luxury_table(df, output_file="edss_agreement.png"):
plt.title('EDSS Subcategory Consistency Analysis', fontsize=16, pad=40, fontweight='bold', color='#2c3e50')
# Add a subtle footer
plt.figtext(0.5, 0.02, "Tolerance: ±0.5 points | N = [Total Samples]",
plt.figtext(0.5, 0.0, "Tolerance: ±0.5 points",
wrap=True, horizontalalignment='center', fontsize=10, color='gray', style='italic')
# Save with high resolution
@@ -561,8 +561,8 @@ create_luxury_table(match_df)
save_styled_table(match_df)
# 6. Save as SVG
#plt.savefig("agreement_table.svg", format='svg', dpi=300, bbox_inches='tight')
#print("Successfully saved agreement_table.svg")
plt.savefig("agreement_table.svg", format='svg', dpi=300, bbox_inches='tight')
print("Successfully saved agreement_table.svg")
# Show plot if running in a GUI environment
plt.show()