Manoj LM

Highly motivated Data Science Intern ready to thrive in demanding digital intelligence processing environments. Well-informed on machine learning and deep learning. Ready to combine tireless hunger for new skills with desire to exploit cutting-edge data science technology.

1. Resume Classification- (NLP Text analysis)- Ongoing

• The Business objective is the document classification solution should significantly reduce the manual human effort in HRM.

2. Forecasting Gold Price in Feature- (Time series forecasting project)

• The objective is to understand the underlying structure in our previous year gold price dataset and come up with a suitable forecasting price.
• Data is already given by client side, did some appropriate steps for to get insights of data set and we understood one thing is nothing but the growth in price has much more of an exponential growth and then we performed the life cycle of data scientist and at the model building stage, we applied some forecast methods are Naïve Forecast, Simple Moving Average and Smoothing methods using accuracy measurements are RMSE and MAPE.
• But after comparing all models we got better accuracy in the ‘Simple Exponential Smoothing' model so we concluded that one is best model for this forecasting.

3. Telecommunication Churn Prediction- (Classification project)

• The aim of this project is to build a feature where customer details will be provided and the model will predict whether the customer is going to churn or not.
• So, we started from data understanding of features are night, Day, Evening charges, international roaming charges etc. and after we understood one thing that is those who have no international plans they are high churners, and we also did feature engineering on certain conditions.
• we used some models they are Log regression, DT, Naïve Bayes, XG Boost, Random Forest But after comparing all models we got better classification through XG Boost so we deployed using this XG Boost in streamlet.

4. Multiclass Image Classification - (CNN-Image Recognition Project)

• The business problem of this project is to rectify the specified my family member image. Total 127 images of 7 members are there which are numbered from 0 to 6.
• Imported what are all required libraries and used Kara's library to load each layer that works on top of TensorFlow and used Image Augmentation Technique for to import the dataset, after that just moved to model building , there performed two process one is building the architecture and model compilation.
• Used Sequential model to build the architecture with activation function rall, because it works very well on all time and then I trained the model with 10 iterations and then got 81% accuracy.