Business Objective

The business goal is predict the strength of the concrete based on several inputs so that we can make the strongest concrete =  More $.

​

Data Science Task

We will perform a neural network analysis on the variables 

​

Select Data

We will use 1030 observations with 8 inputs and 1 output

​

Data Analysis

​

We applied the following steps in our analysis.

1. We performed an multiple regression with a correlation to the true data of .75

2. To improve the analysis we created a NN of 1 hidden layer with 1 node with a correlation to the true data of .80 

3. To improve the NN we changed the hidden layer from 1 node to 5 nodes with a correlation to the true data of .924 

4. To improve the NN we added another hidden layer of 5 nodes with a correlation to the true data of .935

​

Apply Analysis

Her is the original strength, the predicted strength, and the error. 

​

    actual      pred pred_new          error

774  30.14 0.2860639 23.62888 -6.511121

775  44.40 0.4777305 39.46054 -4.939464

776  24.50 0.2840964 23.46636 -1.033635

 

Deploy Model

We would run an experiment for the next month, 50% of jobs will be using old composition and 50% will use the new improved

composition. Distribution of assignment of the composition should be blind.

​

Assess Results

We will evaluate the strength measures of the old vs new and see if new was statistically better. 

​

Strengths of XYZ Analysis

NN has many strengths: Approximate any function, become accurate. 

Weaknesses: Needs lots of data, takes a lot of computing time, takes a lot of effort to create the structure