{"id":46645,"date":"2021-05-11T00:00:00","date_gmt":"2021-05-11T07:00:00","guid":{"rendered":"https:\/\/griddb-linux-hte8hndjf8cka8ht.westus-01.azurewebsites.net\/blog\/bank-loan-classification\/"},"modified":"2025-11-13T12:55:21","modified_gmt":"2025-11-13T20:55:21","slug":"bank-loan-classification","status":"publish","type":"post","link":"https:\/\/griddb-linux-hte8hndjf8cka8ht.westus-01.azurewebsites.net\/en\/blog\/bank-loan-classification\/","title":{"rendered":"Bank Loan Classification"},"content":{"rendered":"

Introduction<\/h2>\n

Today, we will be building a Bank Loan Classification model from scratch using the data stored in GridDB. In this post, we will cover the following:<\/p>\n

1. Storing the data in GridDB <\/a>
\n2. Extracting the data from GridDB<\/a>
\n3. Building a Logistic Regression Model using Pandas<\/a>
\n4. Evaluating our model using heat map and correlation matrix <\/a><\/p>\n

We will begin with installing the prerequisites and setting up our environment. We will be using GridDB’s Python connector<\/a> for this tutorial as the primary language used to model building is Python. GridDB has an extensive range of connecting libraries and APIs<\/a> which makes it easier to deal with both SQL and NoSQL interface.<\/p>\n

Environment and Prerequisites<\/h2>\n

The following tutorial is carried out on Ubuntu Operating system (v. 18.04) with gcc version 7.5.0., GridDB (v. 4.5.2) and Jupyter Notebooks (Anaconda Navigator). As the code is in Python 3, it can be executed on any device\/code editor operating on the said programming language. In case, you are new to Python and Machine Learning, the following links would help you with the installation:<\/p>\n

    \n
  1. Installing Python<\/a><\/li>\n
  2. Installing Anaconda<\/a><\/li>\n<\/ol>\n

    GridDB can be installed using their comprehensive guide available on Github<\/a>. Prerequisites for GridDB’s python-connector<\/a> include: 1. GridDB C-client<\/a> 2. SWIG<\/a> 3. pcre<\/p>\n

    Tutorial on Python-client installation can be found here<\/a><\/p>\n

    Dataset<\/h2>\n

    For this tutorial, we will be working with a Bank Loan Classification dataset which is publicly available on Kaggle<\/a>. There are a total of 5000 instances in the dataset along with 14 attributes. The attributes signify user data evaluated on various criteria such as income, age, experience, etc. The response variable, in this case, is the \u00e2\u20ac\u02dcPersonal Loan\u00e2\u20ac\u2122 variable which is binary in nature. A label of 0 implies a rejection of the loan application whereas 1 conveys an acceptance.<\/p>\n

    The objective is to classify an instance in either of those categories – 0 or 1 (rejected or accepted) based on the rest of the explanatory variables. This is a two-class classification task that could be accomplished using a Logistic Regression Model. So let\u00e2\u20ac\u2122s dive right in!<\/p>\n

    Set the necessary paths<\/h2>\n

    The following code needs to be executed in the Ubuntu Terminal or whichever Operating System you are using after you have installed GridDB<\/a> and the pre-requisites mentioned above.<\/p>\n

    export CPATH=$CPATH:<Python header file directory path>\n\nexport PYTHONPATH=$PYTHONPATH:<installed directory path>\n\nexport LIBRARY_PATH=$LIBRARY_PATH:C client library file directory path\n<\/code><\/pre>\n
    Importing necessary libraries<\/h2>\n
    The first step is, of course, importing the necessary libraries. The libraries we will be using for this tutorial are as follows:<\/p>\n
      \n
    1. NumPy<\/a><\/li>\n
    2. Pandas<\/a><\/li>\n
    3. Matplotlib<\/a><\/li>\n
    4. scikit-learn<\/a><\/li>\n
    5. griddb_python<\/a><\/li>\n<\/ol>\n
      Note that you might encounter an installation error if you’re using these libraries for the first time. Try pip install package-name<\/code> or conda install package-name<\/code>.<\/p>\n
      Once the installation of these libraries is complete, we shall now import the griddb_python<\/code> library. You could run the following command individually on the python console or create a .py file at your convenience. If the installation is successful, the import command should not return any error.<\/p>\n
      import griddb_python\n<\/code><\/pre>\n
      Troubleshooting<\/strong>: It is possible that the build might not be successful. Run the make<\/code> command after setting up the paths. You could also run a sample program<\/a> to figure out the specifics.<\/p>\n
      Adding data to GridDB<\/span><\/h2>\n
      \n
      import griddb_python as griddb\nimport pandas as pd\n\nfactory = griddb.StoreFactory.get_instance()\n\n# Container Initialization\ntry:\n    gridstore = factory.get_store(host=your_host, port=ypur_port, \n            cluster_name=your_cluster_name, username=your_username, \n            password=your_password)\n\n    conInfo = griddb.ContainerInfo(\"Dataset_Name\",\n                    [[\"attribute1\", griddb.Type.INTEGER],[\"attribute2\",griddb.Type.FLOAT],\n                    ....],\n                    griddb.ContainerType.COLLECTION, True)\n    \n    cont = gridstore.put_container(conInfo)   \n    cont.create_index(\"id\", griddb.IndexType.DEFAULT)\n    \n    dataset = pd.read_csv(\"BankLoanClassification.csv\")\n    \n    #Adding data to container\n    for i in range(len(dataset)):\n        ret = cont.put(data.iloc[i, :])\n    print(\"Data has been added\")\n\nexcept griddb.GSException as e:\n    for i in range(e.get_error_stack_size()):\n        print(\"[\", i, \"]\")\n        print(e.get_error_code(i))\n        print(e.get_location(i))\n        print(e.get_message(i))<\/code><\/pre>\n<\/div>\n
      Things to note<\/strong> 1. Replace the parameters of factory.get_store()<\/code> function with your cluster credentials. Alternatively, see<\/a> how to create a new cluster 2. Pass the relevant attribute name, dataset name and, data types to the function griddb.ContainerInfo()<\/code><\/p>\n
      Accessing data from GridDB using SQL<\/span><\/h2>\n
      The griddb_python<\/code> library allows the user to access the data via SQL in Python. This helps us use both languages to our advantage. We can now access what is stored in the database simply by passing a query within our python file like<\/p>\n
      statement = ('SELECT * FROM Dataset_Name')\nsql_query = pd.read_sql_query(statement, cont)\n<\/code><\/pre>\n
      Note that the pd.read_sql_query()<\/code> function converts the data from the SQL query into a pandas dataframe. You can then directly work on the dataframe and build your machine learning model. Alternatively, you can import a csv file directly like:<\/p>\n
      \n
      import os\nimport numpy as np    \nimport pandas as pd   \nimport matplotlib.pyplot as plt    \nfrom sklearn.metrics import confusion_matrix,mean_squared_error,accuracy_score\nfrom sklearn.preprocessing import MinMaxScaler\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.linear_model import LogisticRegression\nfrom sklearn.metrics import accuracy_score \nimport seaborn as sns\nimport matplotlib.pyplot as plt \n<\/code><\/pre>\n<\/div>\n
      \n
      APP_PATH = os.getcwd()\nAPP_PATH\n<\/code><\/pre>\n<\/div>\n
      'C:\\Users\\SHRIPRIYA\\Desktop\\GridDB'\n<\/code><\/pre>\n
      Loading our dataset<\/span><\/h2>\n
      Note that if you have loaded your dataset using GridDB’s python_client<\/code>, you can skip this step as it is redundant.<\/p>\n
      The APP_PATH<\/code> variable contains the current directory which is appended to the file name in the below command. To avoid any complications, make sure that the python file and the csv file are in the same directory. If the case is otherwise, provide the full path of the dataset file to avoid FileNotFoundError<\/code><\/p>\n
      \n
      dataset = pd.read_csv(os.path.join(APP_PATH, 'UniversalBank.csv'))<\/code><\/pre>\n<\/div>\n
      Exploring the dataset<\/h2>\n
      We will be executing some trivial commands to get an overview of our dataset before building a model. Preprocessing and Analysis of your data is a good practice to ensure a more robust and effective model. The head<\/code> command returns the first 5 rows of the dataset. In case, you want to display more, pass a number as an argument to the head<\/code> command. For instance, dataset.head(20)<\/code> would result in the first 20 rows whereas dataset.head(-5)<\/code> would return the last 5.<\/p>\n
      \n
      dataset.head()\n<\/code><\/pre>\n<\/div>\n
      \n