Data Scientist
Department of Health Abu Dhabi
Total years of experience :7 years, 0 Months
Project Name : Marketing Performance Analysis
The Company had its Marketing Team which spends a Budget on advertisement like PPCs, SEOs, Social Media Campaigns and other paid services which create leads to further convert them to clients. I was responsible to collect the data such as amount of Budget spend on different campaigns and the leads generated by them through different sources from Google Analytics, Ad Campaign & CRM. I used to create data tables with various variables to do analysis to find out in which month how much should we spend to get maximum leads with minimum loss and used machine learning regression & deep learning techniques to drive out correlations and predict outcomes such as leads quantity, thus give an idea to the management to plan budget accordingly. I also check the seasonality and trend of the sales using Time Series Analysis & Forecasting using ARIMA model & LSTM, the lead generated by marketing team and its productivity which helps the management to understand at which month we should spend more or less, which further used to prepare a Dashboard for better representation of outcomes achieved.
Project Name : Sales Analysis, Customer Segmentation and CLV
Extract data from CRM and prepare a dataset which shows the behaviour of clients with other variables along with the target variable with Win or Lose, this historical data is been cleaned, pre-process like standardization, normalisation, feature selection, multi-collinear problem, balancing dataset etc then used to train the model for predicting any current client probability of Win or Lose. A Machine Learning Classification Model is been prepared on historical data such as Logistic Regression, Decision Tree, Random Forest, GBM, SVM, Neural Network etc along with the Deep Learning CNN. As per the behaviour of clients which is been transformed to customers, customer segmentation is been done using RFM Modelling & Clustering to find category on the basis of loyalty of customers such as Platinum, Gold, Silver and Bronze. Moreover, Customer Lifetime Value (CLV) is been identified for next few months and helps to decide customer priority and give some offers who are likely to churn.
Project Name : PPC Campaign Analysis
The Marketing which has to spend a good amount of budget on PPC Campaign and they need to choose a right and productive Campaign. In order to select right Campaign historical data is been extracted from CRM & Google Analytics which been cleaned & pre processed to form a proper dataset for analysis with KPIs. The PPC Campaign is categorised into 3 different category as per the KPIs performance and then current data is used to predict each campaign performance in coming month using machine learning algorithms such as K-means Clustering, Hierarchical Clustering, Density based Clustering etc to categorise.
Project Name : Google Ad Words & Keywords
My role is to collect the data from Google Ad Words & Google Analytics account including necessary KPIs for analysis. These KPIs helps to extract the right keywords for each Campaign which gives productive results. These Keywords for each Campaign is been defined into 3 categories and been used place in a Dashboard for decision making.
• Project Name : Sentiment Analysis
With Social Media and Reviews Platform, we extract the comments and chat to perform NLP process using NLTK, Spacy, TextBlob etc. to process the text & understand the sentiments of audience, and also to compare with competitors page. We categorized the comments into 3 categories neural, negative and positive. I used to prepare the dataset of text form and apply various cleaning, lemmatization, removing stop words & punctuation, spelling correction, tokenization, word2vec, word embedding process to prepare the dataset ready for applying classification algorithms using historical data to train the model to predict sentiments.
Project Name : Customer Churn Prediction, Segmentation & CLV
The obtained data from the client’s database is been extracted, cleaned and pre-processed with data mining steps. The website of the client’s Food Company also contains important data like votes, feedback, Satisfaction level, complains, Reviews etc is been extracted by Web Scrapping process and dataset is been prepared for Sentiment Analysis. Customer Segmentation is been done by RFM Modelling and K-Means Clustering to get customer category on the basis of their behaviour such as purchasing frequency, recently and amount of money spend. We also predict which customer will spends how much money in coming few months using CLV method by using models BG/NBD and Gamma-Gamma. After all these models, our aim was to predict which customers are likely to churn in a certain segment of customers, the Model fitting is been done to find the customer churn prediction under certain significance level as per the client’s demand using different machine learning classification algorithms such as Logistic Regression, Decision Tree, Random Forest, GBM, XGBoost, SVM, Deep Learning Classification etc to obtain maximum accuracy and then fresh data is used for churn prediction. The Customers were segregated in 3 zones (red, amber, green) based on their risk score along with the most important feature responsible for churn The result is been visually represented on Dashboard including the data
analysis on scrapped data from website which helps client to get proper picture of customer behaviour & their position in market.
Project Name : Finance and Credit Loan Analysis
We get the historical data along with new data of customer from the client to analyze whom to give credit loan or not on the basis of various parameters of the personal information of customer. We clean, pre-process, missing data handling, feature selection, standardization, data balancing and other data mining processes, then we do multicollinearity check to reveal the co-relation among the variables and then use data visualization to extract patterns & insights which helps to acknowledge customer behaviour. The machine learning algorithms of classification is applied to train a part of historical data and then run the model on the testing data to check the accuracy of model by confusion matrix, ROCR & AUC curve. We apply classification algorithms like Logistic Regression, Decision Tree, Random Forest, Naive Bayes, SVM, Deep Learning Classification etc to obtain maximum accuracy, Precision, Recall and F1 Score from which we select best algorithm.
Project Name : Supermarket Business & Customer Analysis
Cross-selling of different products by testing correlation among them & up-selling product by minimum expense. The Market Basket Analysis for analyzing the correlation among the products sell and evaluating the model by extraction of support, confidence & lift factors. We use Customer Lifetime Value Modelling to know customer loyalty and predict which customer will give how much business in coming few months, we also perform RFM Modelling to do customer segmentation using clustering algorithms. We use to predict the sales and profits in client’s business by using LSTM Deep Learning & ARIMA model for next 3 to 5 years and represent all the reports in different charts and graphs placed in a proper dashboard.
Project Name : Bahanes Food Product Business Analysis
The customer data is used to perform Customer Lifetime Value Modelling to predict which customer will give more business and customer segmentation using clustering algorithms and hence extract high priority customers. We also extract which product has high demand in which region using Exploratory Data Analysis. We forecast the sales using ARIMA, ETS and Prophet Package for next 3 years and placed all report on dashboard.
Project Name : ACT (automatic call testing)
AT&T is holding a typical network testing solution called ACT (automatic call testing) to perform some primary test cases on the Telecom Network. Our responsibility was to perform various data analysis on the generated logs to identify the facts & issues in the network. We use to work on the data which is been provided by the client segregating the meta-data into further different section on the basis of network cluster or site basis. We use to clean and pre-process the data to find the relations between telecom network parameters and extract the clusters having issue in network and represent them statistically how much sites are performing well and where we need to work on to sort out the issues. By extracting the best, good & poor network serving cluster or site on the basis of the log data. We further script on each big data table, the reasons for such issues and how to solve it. We also represent the data visually by using such tools after the analysis on datasets which have been divided on the basis of performance and shared to clients with proper Dashboards.
Project Name : Cluster Drive Testing
AT&T was responsible to perform drive test in a telecom network cluster, which generates a huge amount of data (drive test logs) & consist all the information & meta-information of the network. We were responsible to perform a post analysis on those data & identify whether cluster is working well or not. We use get the data cluster wise to deal with telecom network parameters of 2G, 3G & LTE and perform analysis by cleaning, pre-processing, mining & sorting in the order of parameters which helps to understand the network performance.
Project Name : Sales & Profit Analysis for Indigo Airlines
The Indigo Airlines has a huge business and the client wants to know the status of flights having volume of passengers they are getting in each different cities. The client wants analysis on how many flights each week they should provide on certain destination so in order to reduce costing and earn maximum profit. They want to analyse in each city what is the passengers preferred destination, how to manage flights so that less vacant seats should be there and what is the seasonality and trend on each airport of passenger’s volume. They also want prediction for quantity of passengers in coming next 3 to 5 years. We get the historical data and we perform statistical analysis to find out most preferred destinations by passengers in each city, depending upon this we perform clustering to find out groups for some destination and their volume of passengers. We use ARIMA model to forecast volume of passengers in coming few years monthly and the seasonality and trend according to that client can arrange flights quantity in each city and hence can reduce extra investment & loss and earn maximum profit. All the reports are visually placed on Dashboard hence give clear picture instead of understanding messy data.
Project Name : Analysis on Medical Reports
The clients provides the complete report of the different patients on which they did treatment, we use to work on these datasets by cleaning and applying all data mining and segregate the datasets on the basis of the disease by which patient was suffering. Each dataset is been used for analysis on various parameters which test of patients provide report and examine with the final consequences of treatment whether that person was having that particular disease of not. Using historical data and the client provides the current data of their patients to use machine learning algorithm for classification of patients whether they are suffering from that disease or not.
• CLUSTER DRIVE TEST of 2G, 3G, 4G & VO-LTE NETWORK.
• DETECTING SWAP and rectify the OFC cables. Determining RRU & ANTENNA for required network for optimization requirement & Troubleshooting.
• Checking & Improving the KPI of 2G, 3G and 4G NETWORK.
• Improving in Key Performance Indicators (KPI) like Call Drops, Call blocks, Congestion levels
• To check carrier aggregation in 4G LTE NETWORK and obtained bandwidth up to 100 MHz
• Making reports on MAP INFO SOFTWARE to check cluster drive DOWLOADING/UPLOADING SPEED using THEMATIC MAP with proper LEGEND.
• Manipulation of parameters in 4G Data obtained from TEMS Logs using Excel, Advance Excel & SQL to ANALYZE the network quality in different CLUSTERS and making network reports.
Telecommunication & Wireless Technology