Sequential_Logistics

A sequential logistics model for predicting the label, which is salary of differents adults based on their 'age','workclass','education',etc.

Getting started

Use the following command to clone the repository on to your system

git clone https://github.com/vini7148/Sequential_Logistics.git
cd Seqential_Logistics

Prerequisites

An environment for running .ipynb files is required along with the following python libraries

pandas
matplotlib
seaborn
sklearn
tensorflow-gpu
tensorflow.keras

Running the model

After acquiring all the required pyhton libraries and an envitronment for executing .ipynb files along with this model. Open the "Adults using Sequential Logistic.ipynb" file

Note: if you don't have tensorflow-gpu or cannot use a discrete GPU then please comment out the 'with tf.device("/device:GPU:1"):' in the executable cell numbered 12 of the .ipynb file and correct the indentation of the following lines of code for it to run successfully.

Built with

• Python 3
• Anaconda
• Jupyter notebookconda install -c conda-forge jupyterlab

Note if you have not install Anaconda, you can still install Jupyter Notebook by pip install jupyterlab

Description of the keras.sequential model

The activation function used is

activation = tf.nn.sigmoid

The optimizer used is

optimizer = tf.keras.optimizers.Adam(0.01)

The summary of the model is

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 64)                3712      
_________________________________________________________________
dense_1 (Dense)              (None, 64)                4160      
_________________________________________________________________
dense_2 (Dense)              (None, 64)                4160      
_________________________________________________________________
dense_3 (Dense)              (None, 1)                 65        
=================================================================
Total params: 12,097
Trainable params: 12,097
Non-trainable params: 0
_________________________________________________________________

Training of the model

This sequential model was trained for 100 epochs

Note: if you are not using tensorflow-gpu, you might want to decrease the number of epochs as it will take a long time to train on CPU. Use 20 epochs for effective results on CPU.

Outcome of this model

For the last epoch the loss, mean absolute error, etc. is

   loss     mean_absolute_error mean_squared_error val_loss val_mean_absolute_error val_mean_squared_error epoch
99 0.183517 0.366432            0.183517           0.186422 0.363392                0.186422               99

The confusion matrix for the test dataset("Adults_Test.csv") is as follows

[[11360     0]
 [ 3609    91]]