final_analysis

zf_task_code_submit.ipynb

@@ -868,7 +868,14 @@
  "cell_type": "markdown",
  "metadata": {},
  "source": [
- "![age](https://raw.githubusercontent.com/vanshilshah97/pred_maint_zf/main/images/pivoting_on_age2.png)"
+ "![failure_model_ratio](https://raw.githubusercontent.com/vanshilshah97/pred_maint_zf/main/images/failure_model_ratio.png)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "#### We can see that model2 variant of the machine is prone to more errors, and this is further highlighted in the case when I find the feature coefficient in logistic regression it has the second most and positive correlation with the probabilty of a machine being unhealthy "
  ]
  },
  {
@@ -3215,7 +3222,7 @@
  "cell_type": "markdown",
  "metadata": {},
  "source": [
- "#### This is the feature importance plot generated by the random forest. It marks age, one-hot encoding and rotation as some of the important features in capturing the anomalies. I believe this information could used in our future work while working in an unsupervised domain for reducing feature dimensionality"
+ "### This is the feature importance plot generated by the random forest. It marks age, one-hot encoding and rotation as some of the important features in capturing the anomalies. I believe this information could used in our future work while working in an unsupervised domain for reducing feature dimensionality"
  ]
  },
  {
@@ -3243,8 +3250,8 @@
  "cell_type": "markdown",
  "metadata": {},
  "source": [
- "#### As we can finally see that random forest gives superior performance, we can furhter imporve the recall by doing a gridsearch on various hyperparameters of random forest . Having a comparable performance are boosting techniques, I think with proper fine tuning of the parameters of gradient boosting techniques we could also improve the recall significantly.\n",
- "#### Hence these methods are the way to go"
+ "### As we can finally see that random forest gives superior performance, we can furhter imporve the recall by doing a gridsearch on various hyperparameters of random forest . Having a comparable performance are boosting techniques, I think with proper fine tuning of the parameters of gradient boosting techniques we could also improve the recall significantly.\n",
+ "### Hence these methods are the way to go"
  ]
  },
  {