Develop Time Series Web service in Python. Web service divided in three modules data analyse, prepared the arima model and predict the forecast. Design of the first module for trend, seasonal and residual. In second module base on data nature automatically generate the arima model . and third phase predicate forecast value using arima which develop in second phase.

Model: Arima ,
Technique : Time series with Forecasting
Package: Falsk, arima, pandas, numpy, pickle
Language : python

Time series purely work base on web service which is develop in python using flask package. ( Consider as API)

The ESC-50 dataset is a labeled collection of 2000 environmental audio recordings suitable for bench marking methods of environmental sound classification. The dataset consists of 5-second-long recordings organized into 50 semantical classes (with 40 examples per class) loosely arranged into 5 major categories. I trained Convolution Neural Network for sound classification. I achieved classification accuracy of approx ~83%. MFCC (mel-frequency cepstrum) feature is used to train models. Other features like short term fourier transform, chroma, melspectrogram can also be extracted. This dataset is challenged with 50 classes related to accuracy of Audio prediction. Achieved 83?curacy with mapping weight technique of previous maturity of model. I got prediction 80% correct.

The largest memory manufacturer wants to classify defective and non-defective images with high accuracy and a lower overkill rate.

In this project, the client provided 10K images containing defective and non-defective images. The main objective of the project is to classify the two type of images and store the result. Project required to read massive images of the network path defined in the configuration. We have developed system which runs every hour and reads the files from the network. These files are passed into the trained model to predict the result. The results are saved in two formats, one in a flat file and the other in the database, which can be used to generate reports.

The challenge was to identify defective images with a high accuracy of nearly 99%. As a projectile, we need more images to form the deep learning model generated by the image enhancement technique. In a second step, read the image using Open CV and apply the learning transfer to develop the model.

As a result, our image classification model obtained a satisfactory default image classification with a Type 2 error of zero.

Responsibilities:

• Collection of requirements of the Business team.
• Increase training data using the image augmentation technique.
• Used the concept of transfer learning to obtain good accuracy.
• Implementation of the Deep Learning Image Classification Model.
• Model validated in thousands of images.
• Implementation of the model in GPU.
• Weekly Scrum meeting.

The ubiquity of information today empowers investors at any scale to make better investment decisions. Challenge in this project to ingesting and interpreting the data to figure out which information is helpful, finding the flag in this ocean of data. By investigating news data to anticipate(predict) stock prices, and the condition of research in understanding the prescient power of the news. harnessed, to help predict financial outcomes and produce critical monetary effect everywhere throughout the world. This undertaking depends on news data affect so we dissect the sentiments first using TFIDF(NLP Technique) and other NLP techniques which generate features and check their effect on stocks.

Responsibilities:

• Requirement gathering from the Business Team.
• Gather market and news data.
• Explore data to understand the sentiments.
• Generate feature of news data using TFIDF.
• Implemented and Designed different parameter for ML model.(Microsoft Designed Light GBM)
• Model validation/Deployment

In this project one of leading automotive manufacture wants to detect When their robots need a maintenance before shutdown. So based on problem they can save their build an LSTM network in order to predict remaining useful life (or time to failure) of robots. The network uses simulated robot’s sensor values to predict when an robot will fail in the future so that maintenance can be planned in advance.

Time Series Analysis : How many more cycles an in-service robot will last before it fails?
Binary classification: Is this robot going to fail within number of cycles?

Responsibilities:
• Requirement gathering from the Business Team.
• Gathering data.
• Used sequence to sequence Modeling approach.
• Training LSTM model.
• Model validation/Deployment.

Real estate firm wants property recommendation to dealers. Client has large number of property data and wanted recommendation system to their dealers. Data is sparse, so It suffer from scalability issue which is solved using cluster approach and new listed property doesn’t get much attention which is cold start problem and solved by designing hybrid similarity approach.

Responsibilities:

• Requirement gathering from the Business Team.
• Matrix Generation from raw data.
• Identify and solved cold-start problem
• Solved sparse data issue using clustering approach.
• Implemented and Designed different similarity algorithm and recommendation.
• Model validation/Deployment.

Platform support all kind of data-set. Build the Dynamic functionality to change the parameter and function. All kind of algorithm develop for classification and regression. Dynamic model save and reuse this model for prediction.

Algorithm Name: SVM, Neural Network, Decision Tree, K-NN, Linear Regression, Gradient Boosting, and so on.....
Technique : Classification & Regression
Language : Python
Package : Numpy, Pandas, Sklearn, Flask, Pickle,

Major challenge was different project wise prepared model,find the forecasting value and plot the time series values in power bi.

Model: Auto-Arima
Technique : Time series with Forecasting
Package: tsseries, forecast, autoarima, dplyr
Tools: R and Power Bi

Mainly the purpose of the project creates a search engine based on natural language processing and machine learning. On user search query bot give the answer relevant his question.

Language : Python
Technique : NLP with Machine Learning, Ranking Algorithm
Package : Numpy, Pandas, NLTK, Flask, Pdf-Miner, Scipy

Project base on housing price prediction. main challenge ware feature selection and prepared the model using regression technique. project aim was to give price area wise and consider all parameter like location, power station, garden, recreation area/fun park, job distance,mall etc..

Model: Random Forest , Gradient Boosting Algorithm
Technique : Regression
Package: RandomForest, xgb, plotly, shiny, dplyr
Language : R