My Leaderboard

Secured 301 Position out of 2144 participants with Prediction Score of: 85.17497

The Process is divided in 4 Steps

• Cleaning the Data
• Removing Outliers
• EDA on cleaned dataset
• Model Building

Cleaning The Data-Set

Filling NaN values with Median for Train Dataset

fill_na_train = train_df['Stall_no'].median()
train_df['Stall_no'].fillna(fill_na_train, inplace=True)
fill_na_train = train_df['Discount_avail'].median()
train_df['Discount_avail'].fillna(fill_na_train, inplace=True)
fill_na_train = train_df['charges_1'].median()
train_df['charges_1'].fillna(fill_na_train, inplace=True)
fill_na_train = train_df['charges_2 (%)'].median()
train_df['charges_2 (%)'].fillna(fill_na_train, inplace=True)
fill_na_train = train_df['Minimum_price'].median()
train_df['Minimum_price'].fillna(fill_na_train, inplace=True)
fill_na_train = train_df['Maximum_price'].median()
train_df['Maximum_price'].fillna(fill_na_train, inplace=True)
fill_na_train = train_df['Selling_Price'].median()
train_df['Selling_Price'].fillna(fill_na_train, inplace=True)

Checking all values were filled

train_df.isnull().sum()

Output:
```Product_id 0
Stall_no 0
instock_date 0
Market_Category 0
Customer_name 211
Loyalty_customer 0
Product_Category 0
Grade 0
Demand 0
Discount_avail 0
charges_1 0
charges_2 (%) 0
Minimum_price 0
Maximum_price 0
Selling_Price 0
dtype: int64

**Filling NaN values with Median for Test Dataset**
```python
fill_na_test = test_df['Stall_no'].median()
test_df['Stall_no'].fillna(fill_na_test, inplace=True)
fill_na_test = test_df['charges_1'].median()
test_df['charges_1'].fillna(fill_na_test, inplace=True)
fill_na_test = test_df['charges_2 (%)'].median()
test_df['charges_2 (%)'].fillna(fill_na_test, inplace=True)
fill_na_test = test_df['Minimum_price'].median()
test_df['Minimum_price'].fillna(fill_na_test, inplace=True)

Checking all values were filled

test_df.isnull().sum()

Output:

Product_id           0
Stall_no             0
instock_date         0
Market_Category      0
Customer_name       53
Loyalty_customer     0
Product_Category     0
Grade                0
Demand               0
Discount_avail       0
charges_1            0
charges_2 (%)        0
Minimum_price        0
Maximum_price        0
dtype: int64

Removing The Outlier

Removed Outliers by Using IQR method:

def outlier_removal(column):
    sorted(column)
    Q1, Q3 = np.percentile(column, [25, 75])
    IQR = Q3 - Q1
    lower_value = Q1 - (1.5 * IQR)
    upper_value = Q3 + (1.5 * IQR)
    return lower_value, upper_value

Doing Some EDA on the cleaned dataset.

Plotting Distribution for the Data-Set

1. Selling Price Distribution

2. Charges_1 Distribution

3. Maximum_Price Distribution

4. Product Sold Per Category

5. Products Sold As Per Loyality Customers

6. Products Sold Per Year(2014, 2015, 2016)

7. Loyality Customers

Model Building

1. Model Preparation

   Data for X:

   train[['Stall_no', 'Market_Category', 'Grade', 'Demand', 'Discount_avail', 'charges_1', 'charges_2 (%)', 'Minimum_price', 'Maximum_price', 'month', 'year', 'day', 'loyality_customers', 'Product_Category_Cosmetics', 'Product_Category_Educational', 'Product_Category_Fashion', 'Product_Category_Home_decor', 'Product_Category_Hospitality', 'Product_Category_Organic', 'Product_Category_Pet_care', 'Product_Category_Repair', 'Product_Category_Technology']]
   
   

   Data for Y:

   train['Selling_Price']

1. Train_Test_Split

    from sklearn.model_selection import train_test_split
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=10)

   
   

2. Using Random Forest

    from sklearn.ensemble import RandomForestRegressor
    reg = RandomForestRegressor(max_depth = 15, random_state=0, n_estimators = 100, verbose = 1)
    reg.fit(X_train,y_train)

    reg.score(X_train, y_train)

   Reg.score: 0.996330144542303
   
   
   

• R^2 Value

  Value for rsquare is 0.9722337968809527
  
  

• RMSE Value

  Value for Root Mean Square Error is : 401.4823455561136
  
  

• RMSE Value

  Value for Root Mean Square Error is : 401.4823455561136

Hope You Liked My Competition Walkthrough :)

😀 😄 😃