Update app.py

app.py

@@ -25,21 +25,21 @@ nurse_actions = ["Monitor Vitals", "Administer Medication", "Report to Doctor",
 
 # Reward and penalty system
 rewards = {
-    "Prescribe Medication": 10,
-    "Recommend Tests": 5,
-    "Consult Clinician": 7,
-    "Schedule Surgery": 15,
-    "Monitor Vitals": 3,
-    "Administer Medication": 10,
-    "Review Diagnostic Test": 6,
-    "Recommend Additional Tests": 4
+    "Prescribe Medication": 12,  # Slightly increased reward
+    "Recommend Tests": 7,  # Slightly increased reward
+    "Consult Clinician": 9,
+    "Schedule Surgery": 17,
+    "Monitor Vitals": 4,
+    "Administer Medication": 12,
+    "Review Diagnostic Test": 7,
+    "Recommend Additional Tests": 6
 }
 
 penalties = {
-    "Wrong Medication": -10,
-    "Missed Diagnosis": -20,
-    "Incorrect Test Recommendation": -5,
-    "Stress-Induced Mistake": -15
+    "Wrong Medication": -5,  # Reduced penalty
+    "Missed Diagnosis": -10,
+    "Incorrect Test Recommendation": -3,
+    "Stress-Induced Mistake": -7
 }
 
 # Initialize the session state for counting treatments
@@ -58,13 +58,13 @@ class Agent:
         self.q_table = np.zeros((len(patient_conditions), len(doctor_actions if agent_type == "Doctor" else nurse_actions)))
         self.state = random.choice(patient_conditions)
     
-    def choose_action(self, exploration_rate=0.05):  # Reduce exploration rate
+    def choose_action(self, exploration_rate=0.03):  # Further reduce exploration rate
         if random.uniform(0, 1) < exploration_rate:  # Explore
             return random.randint(0, len(doctor_actions)-1 if self.agent_type == "Doctor" else len(nurse_actions)-1)
         else:  # Exploit (choose the best action based on Q-values)
             return np.argmax(self.q_table)
     
-    def update_q_value(self, state, action, reward, learning_rate=0.3, discount_factor=0.95):  # Increase learning rate
+    def update_q_value(self, state, action, reward, learning_rate=0.25, discount_factor=0.95):  # Increase learning rate
         old_q_value = self.q_table[state, action]
         best_future_q_value = np.max(self.q_table)
         new_q_value = old_q_value + learning_rate * (reward + discount_factor * best_future_q_value - old_q_value)
@@ -104,7 +104,11 @@ if st.button("Run Simulation"):
     st.write(f"Nurse's Chosen Action: {nurse_actions[nurse_action]}")
 
     # Complications that can arise during treatment
-    complication = random.choice([None, "Allergic Reaction", "Unexpected Complication"])
+    complication = random.choices(
+        [None, "Allergic Reaction", "Unexpected Complication"],
+        weights=[0.6, 0.2, 0.2]  # 60% chance no complications, 20% for each complication
+    )[0]
+    
     if complication:
         st.subheader(f"Complication: {complication}")
         if complication == "Allergic Reaction":
@@ -137,7 +141,10 @@ if st.button("Run Simulation"):
     st.write(f"Patient Satisfaction: {st.session_state.patient_satisfaction}")
 
     # Simulated patient feedback and next steps
-    next_step = random.choice(["Recovery", "Further Treatment Needed", "Complication"])
+    next_step = random.choices(
+        ["Recovery", "Further Treatment Needed", "Complication"],
+        weights=[0.6, 0.3, 0.1]  # Biased toward successful recovery
+    )[0]
     st.write(f"Patient Status after treatment: {next_step}")
 
     # Increment performance based on outcome