Social media has changed the way people communicate with each other and has brought people togeth... more Social media has changed the way people communicate with each other and has brought people together. Enterprises are increasingly using it as a medium for marketing activities. However, due to the size of these networks, marketers often look for key customers (influencers) to drive the campaign to the community. In this paper, we take a game theoretic approach to identify key influencers in a network. We begin with defining coalition games to model the social network and then use the concept of Banzhaf index to measure the utility of each user to the coalition. We further extend this concept towards identification of influencers and compare the resulting algorithm against existing works on influence maximization on several datasets. Improvements are observed.
Customers are usually exposed to online digital advertisement channels, such as email marketing, ... more Customers are usually exposed to online digital advertisement channels, such as email marketing, display advertising, paid search engine marketing, along their way to purchase or subscribe products( aka. conversion). The marketers track all the customer journey data and try to measure the effectiveness of each advertising channel. The inference about the influence of each channel plays an important role in budget allocation and inventory pricing decisions. Several simplistic rule-based strategies and data-driven algorithmic strategies have been widely used in marketing field, but they do not address the issues, such as channel interaction, time dependency, user characteristics. In this paper, we propose a novel attribution algorithm based on deep learning to assess the impact of each advertising channel. We present Deep Neural Net With Attention multi-touch attribution model (DNAMTA) model in a supervised learning fashion of predicting if a series of events leads to conversion, and ...
Customers are usually exposed to online digital advertisement channels , such as email marketing,... more Customers are usually exposed to online digital advertisement channels , such as email marketing, display advertising, paid search engine marketing, along their way to purchase or subscribe prod-ucts(aka. conversion). The marketers track all the customer journey data and try to measure the effectiveness of each advertising channel. The inference about the influence of each channel plays an important role in budget allocation and inventory pricing decisions. Several simplistic rule-based strategies and data-driven al-gorithmic strategies have been widely used in marketing field, but they do not address the issues, such as channel interaction, time dependency, user characteristics. In this paper, we propose a novel attribution algorithm based on deep learning to assess the impact of each advertising channel. We present Deep Neural Net With Attention multi-touch attribution model (DNAMTA) model in a supervised learning fashion of predicting if a series of events leads to conversion, and it leads us to have a deep understanding of the dynamic interaction effects between media channels. DNAMTA also incorporates user-context information, such as user demographics and behavior, as control variables to reduce the estimation biases of media effects. We used computational experiment of large real world marketing dataset to demonstrate that our proposed model is superior to existing methods in both conversion prediction and media channel influence evaluation.
Social media has changed the way people communicate with each other and has brought people togeth... more Social media has changed the way people communicate with each other and has brought people together. Enterprises are increasingly using it as a medium for marketing activities. However, due to the size of these networks, marketers often look for key customers (influencers) to drive the campaign to the community. In this paper, we take a game theoretic approach to identify key influencers in a network. We begin with defining coalition games to model the social network and then use the concept of Banzhaf index to measure the utility of each user to the coalition. We further extend this concept towards identification of influ-encers and compare the resulting algorithm against existing works on influence maximization on several datasets. Improvements are observed.
A dog image dataset, which contains subtle inter-class differences and large intra-class variatio... more A dog image dataset, which contains subtle inter-class differences and large intra-class variations, is collected in [1] to explore the challenges faced in Fine-Grained Image Classification. However, the dataset collected is restricted to pure breed dogs. In reality, mixed breed dogs are much more common and identifying their parent breed is useful. Since there is no mixed breed dog dataset collection known in literature, in this work, we collect a mixed breed dataset and investigate the performance of our classification algorithm. We fine-tune the pre-trained models such as Inception using pure breeds dataset and then use the fine-tuned model to classify mixed breed dogs into their parent breeds. We measure the accuracy of identifying at least one parent and both parents separately in Top K predictions. Further, we improve our model by adding saliency maps and augmented data to our training data and then fine-tune the pre-trained models on entire data.
As a group of 3 people, we worked on tracking Web
Analytics data and deriving behavioral insight... more As a group of 3 people, we worked on tracking Web
Analytics data and deriving behavioral insights of PayPal customers using Hadoop. I implemented a pattern
matching algorithm as map reduce jobs in Java to find out PayPal customer pain points. We wrote complex
logic to group customer pages on data greater than 1 TB using custom Hadoop partitioners. Greater accuracy
achieved for more than 90 percent of the possible paths taken by PayPal's customers.
We study the problem of selling identical goods to
several possible unit-demand bidders in a sett... more We study the problem of selling identical goods to several possible unit-demand bidders in a setting in which the total supply of goods as well as the total number of bidders is unknown to the mechanism. Items arrive and expire dynamically, while the agents arrive and depart dynamically, and the seller must make the allocation and payment decisions online. We consider single valued preference domains and characterize the space of decision policies that can be truthfully implemented in a dominant strategy equilibrium. Finally, we relax the assumption of no early misreports to extend the already existing mechanism design for a setting where the agents arrive and depart dynamically in discrete time periods and a single indivisible item is to be allocated in each discrete time period.
We tried to learn different ways of estimating VaR. The most common ways i.e. historical simulat... more We tried to learn different ways of estimating VaR. The most common ways i.e. historical simulation, variance–covariance method and monte carlo simulations are commonly used. We explored the use of order statistics in VaR estimation. We observed that when the $k^{th}$ heaviest loss is used the expected quantile and its error will be independent of the portfolio composition and the returns functions of the assets in the portfolio. We then analyzed the properties of different VaR-estimators by simulating different return distributions. We show that the predictive power of a VaR estimate at higher confidence intervals with limited amount of data should be used carefully for there is a large chance of error.
Wrote map reduce jobs in python to parse the 25 GB patient data (medical claims) in JSON format. ... more Wrote map reduce jobs in python to parse the 25 GB patient data (medical claims) in JSON format. Clustered and categorized the users based on the features in the data using random forests in sklearn. Built a large scale fraud detection algorithm with 95 percent precision as a Hadoop map reduce program
Social media has changed the way people communicate with each other and has brought people togeth... more Social media has changed the way people communicate with each other and has brought people together. Enterprises are increasingly using it as a medium for marketing activities. However, due to the size of these networks, marketers often look for key customers (influencers) to drive the campaign to the community. In this paper, we take a game theoretic approach to identify key influencers in a network. We begin with defining coalition games to model the social network and then use the concept of Banzhaf index to measure the utility of each user to the coalition. We further extend this concept towards identification of influencers and compare the resulting algorithm against existing works on influence maximization on several datasets. Improvements are observed.
Customers are usually exposed to online digital advertisement channels, such as email marketing, ... more Customers are usually exposed to online digital advertisement channels, such as email marketing, display advertising, paid search engine marketing, along their way to purchase or subscribe products( aka. conversion). The marketers track all the customer journey data and try to measure the effectiveness of each advertising channel. The inference about the influence of each channel plays an important role in budget allocation and inventory pricing decisions. Several simplistic rule-based strategies and data-driven algorithmic strategies have been widely used in marketing field, but they do not address the issues, such as channel interaction, time dependency, user characteristics. In this paper, we propose a novel attribution algorithm based on deep learning to assess the impact of each advertising channel. We present Deep Neural Net With Attention multi-touch attribution model (DNAMTA) model in a supervised learning fashion of predicting if a series of events leads to conversion, and ...
Customers are usually exposed to online digital advertisement channels , such as email marketing,... more Customers are usually exposed to online digital advertisement channels , such as email marketing, display advertising, paid search engine marketing, along their way to purchase or subscribe prod-ucts(aka. conversion). The marketers track all the customer journey data and try to measure the effectiveness of each advertising channel. The inference about the influence of each channel plays an important role in budget allocation and inventory pricing decisions. Several simplistic rule-based strategies and data-driven al-gorithmic strategies have been widely used in marketing field, but they do not address the issues, such as channel interaction, time dependency, user characteristics. In this paper, we propose a novel attribution algorithm based on deep learning to assess the impact of each advertising channel. We present Deep Neural Net With Attention multi-touch attribution model (DNAMTA) model in a supervised learning fashion of predicting if a series of events leads to conversion, and it leads us to have a deep understanding of the dynamic interaction effects between media channels. DNAMTA also incorporates user-context information, such as user demographics and behavior, as control variables to reduce the estimation biases of media effects. We used computational experiment of large real world marketing dataset to demonstrate that our proposed model is superior to existing methods in both conversion prediction and media channel influence evaluation.
Social media has changed the way people communicate with each other and has brought people togeth... more Social media has changed the way people communicate with each other and has brought people together. Enterprises are increasingly using it as a medium for marketing activities. However, due to the size of these networks, marketers often look for key customers (influencers) to drive the campaign to the community. In this paper, we take a game theoretic approach to identify key influencers in a network. We begin with defining coalition games to model the social network and then use the concept of Banzhaf index to measure the utility of each user to the coalition. We further extend this concept towards identification of influ-encers and compare the resulting algorithm against existing works on influence maximization on several datasets. Improvements are observed.
A dog image dataset, which contains subtle inter-class differences and large intra-class variatio... more A dog image dataset, which contains subtle inter-class differences and large intra-class variations, is collected in [1] to explore the challenges faced in Fine-Grained Image Classification. However, the dataset collected is restricted to pure breed dogs. In reality, mixed breed dogs are much more common and identifying their parent breed is useful. Since there is no mixed breed dog dataset collection known in literature, in this work, we collect a mixed breed dataset and investigate the performance of our classification algorithm. We fine-tune the pre-trained models such as Inception using pure breeds dataset and then use the fine-tuned model to classify mixed breed dogs into their parent breeds. We measure the accuracy of identifying at least one parent and both parents separately in Top K predictions. Further, we improve our model by adding saliency maps and augmented data to our training data and then fine-tune the pre-trained models on entire data.
As a group of 3 people, we worked on tracking Web
Analytics data and deriving behavioral insight... more As a group of 3 people, we worked on tracking Web
Analytics data and deriving behavioral insights of PayPal customers using Hadoop. I implemented a pattern
matching algorithm as map reduce jobs in Java to find out PayPal customer pain points. We wrote complex
logic to group customer pages on data greater than 1 TB using custom Hadoop partitioners. Greater accuracy
achieved for more than 90 percent of the possible paths taken by PayPal's customers.
We study the problem of selling identical goods to
several possible unit-demand bidders in a sett... more We study the problem of selling identical goods to several possible unit-demand bidders in a setting in which the total supply of goods as well as the total number of bidders is unknown to the mechanism. Items arrive and expire dynamically, while the agents arrive and depart dynamically, and the seller must make the allocation and payment decisions online. We consider single valued preference domains and characterize the space of decision policies that can be truthfully implemented in a dominant strategy equilibrium. Finally, we relax the assumption of no early misreports to extend the already existing mechanism design for a setting where the agents arrive and depart dynamically in discrete time periods and a single indivisible item is to be allocated in each discrete time period.
We tried to learn different ways of estimating VaR. The most common ways i.e. historical simulat... more We tried to learn different ways of estimating VaR. The most common ways i.e. historical simulation, variance–covariance method and monte carlo simulations are commonly used. We explored the use of order statistics in VaR estimation. We observed that when the $k^{th}$ heaviest loss is used the expected quantile and its error will be independent of the portfolio composition and the returns functions of the assets in the portfolio. We then analyzed the properties of different VaR-estimators by simulating different return distributions. We show that the predictive power of a VaR estimate at higher confidence intervals with limited amount of data should be used carefully for there is a large chance of error.
Wrote map reduce jobs in python to parse the 25 GB patient data (medical claims) in JSON format. ... more Wrote map reduce jobs in python to parse the 25 GB patient data (medical claims) in JSON format. Clustered and categorized the users based on the features in the data using random forests in sklearn. Built a large scale fraud detection algorithm with 95 percent precision as a Hadoop map reduce program
Uploads
Analytics data and deriving behavioral insights of PayPal customers using Hadoop. I implemented a pattern
matching algorithm as map reduce jobs in Java to find out PayPal customer pain points. We wrote complex
logic to group customer pages on data greater than 1 TB using custom Hadoop partitioners. Greater accuracy
achieved for more than 90 percent of the possible paths taken by PayPal's customers.
several possible unit-demand bidders in a setting
in which the total supply of goods as well as the
total number of bidders is unknown to the mechanism.
Items arrive and expire dynamically, while
the agents arrive and depart dynamically, and the
seller must make the allocation and payment decisions
online. We consider single valued preference
domains and characterize the space of decision
policies that can be truthfully implemented in
a dominant strategy equilibrium. Finally, we relax
the assumption of no early misreports to extend
the already existing mechanism design for a setting
where the agents arrive and depart dynamically in
discrete time periods and a single indivisible item
is to be allocated in each discrete time period.
Analytics data and deriving behavioral insights of PayPal customers using Hadoop. I implemented a pattern
matching algorithm as map reduce jobs in Java to find out PayPal customer pain points. We wrote complex
logic to group customer pages on data greater than 1 TB using custom Hadoop partitioners. Greater accuracy
achieved for more than 90 percent of the possible paths taken by PayPal's customers.
several possible unit-demand bidders in a setting
in which the total supply of goods as well as the
total number of bidders is unknown to the mechanism.
Items arrive and expire dynamically, while
the agents arrive and depart dynamically, and the
seller must make the allocation and payment decisions
online. We consider single valued preference
domains and characterize the space of decision
policies that can be truthfully implemented in
a dominant strategy equilibrium. Finally, we relax
the assumption of no early misreports to extend
the already existing mechanism design for a setting
where the agents arrive and depart dynamically in
discrete time periods and a single indivisible item
is to be allocated in each discrete time period.