Sakshi Sharma

# Gas Level Prediction Model (LSTM)
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense

model = Sequential([
    LSTM(64, input_shape=(60, 5)),  # 60 timesteps, 5 sensor features
    Dense(32, activation='relu'),
    Dense(1)  # Predicts gas concentration
])

model.compile(optimizer='adam', loss='mse')
model.fit(X_train, y_train, epochs=50)
                    

# ESP32 Sensor Data Collection
import machine
from machine import Pin, I2C
import time

mq2 = machine.ADC(Pin(34))
dht = dht22.DHT22(Pin(14))

while True:
    gas = mq2.read()
    temp, hum = dht.read()
    data = f"{gas},{temp},{hum}"
    send_via_lora(data)
    time.sleep(1)
                    

# customer segmentation
import pandas as pd
from sklearn.cluster import KMeans
from sklearn.preprocessing import StandardScaler

df1 = NewDf

X = df1[['discounted_price', 'rating_count']]
# X = X.fillna(0)
scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)

kmeans = KMeans(n_clusters=3, random_state=42)
df1['Customer_Segment'] = kmeans.fit_predict(X_scaled)
NewDf['Customer_Segment'] = df1['Customer_Segment']
output = NewDf
output
                    

# sales prediction
import numpy as np

df = NewDf.copy()

if 'sales' not in df.columns:
    df['sales'] = df['rating_count'] * df['discounted_price']

df['log_sales'] = np.log1p(df['sales'])

df['sales'] = df['sales'].astype(float)
df['rating'] = df['rating'].astype(float)
df['rating_count'] = df['rating_count'].astype(float)
                    

# data generation
import pandas as pd
import numpy as np

np.random.seed(42)
n_samples = 5000
data = pd.DataFrame({
'applicant_income': np.random.normal(loc=35000, scale=8000, size=n_samples).clip(5000, 100000),
'claimed_subsidy_amount': np.random.normal(loc=7000, scale=1500, size=n_samples).clip(1000, 15000),
'land_owned_acres': np.random.uniform(0.2, 8.0, size=n_samples),
'number_of_dependents': np.random.randint(0, 7, size=n_samples),
'region': np.random.choice(['North', 'South', 'East', 'West'], size=n_samples),
'previous_claims': np.random.poisson(1.0, size=n_samples),
'is_employed': np.random.choice([0, 1], size=n_samples, p=[0.3, 0.7])
})
                    

# data preprocessing
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder, StandardScaler
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report, confusion_matrix

# Load dataset
df = pd.read_csv("synthetic_subsidy_data.csv")

# Separate features and target
X = df.drop("is_fraud", axis=1)
y = df["is_fraud"]

# One-hot encode categorical variables
categorical_cols = ["subsidy_type", "location"]
encoder = OneHotEncoder(sparse_output=False, drop="first")
X_encoded = encoder.fit_transform(X[categorical_cols])

# Drop original categorical columns and combine encoded features
X_numeric = X.drop(columns=categorical_cols)
X_final = np.concatenate([X_numeric.values, X_encoded], axis=1)
                    

# data generation
import pandas as pd
import numpy as np

np.random.seed(42)
n_samples = 5000
data = pd.DataFrame({
'applicant_income': np.random.normal(loc=35000, scale=8000, size=n_samples).clip(5000, 100000),
'claimed_subsidy_amount': np.random.normal(loc=7000, scale=1500, size=n_samples).clip(1000, 15000),
'land_owned_acres': np.random.uniform(0.2, 8.0, size=n_samples),
'number_of_dependents': np.random.randint(0, 7, size=n_samples),
'region': np.random.choice(['North', 'South', 'East', 'West'], size=n_samples),
'previous_claims': np.random.poisson(1.0, size=n_samples),
'is_employed': np.random.choice([0, 1], size=n_samples, p=[0.3, 0.7])
})
                    

# data preprocessing
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder, StandardScaler
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report, confusion_matrix

# Load dataset
df = pd.read_csv("synthetic_subsidy_data.csv")

# Separate features and target
X = df.drop("is_fraud", axis=1)
y = df["is_fraud"]

# One-hot encode categorical variables
categorical_cols = ["subsidy_type", "location"]
encoder = OneHotEncoder(sparse_output=False, drop="first")
X_encoded = encoder.fit_transform(X[categorical_cols])

# Drop original categorical columns and combine encoded features
X_numeric = X.drop(columns=categorical_cols)
X_final = np.concatenate([X_numeric.values, X_encoded], axis=1)