Social network analysis and growth recommendations for DataScience SG community
1 like1,286 views
This document discusses different social networks and connections between users on a platform. It describes user friendship networks colored based on inactivity levels. It also describes user like networks, with edges representing likes between friends and non-friends. Finally, it discusses user comment networks derived from the like network connections.
1 of 33
Downloaded 21 times
Ad
Recommended
AXA x DSSG Meetup Sharing (Feb 2016)
AXA x DSSG Meetup Sharing (Feb 2016)Eugene Yan Ziyou
油
AXA is leveraging big data through its Data Innovation Lab to transform into a data-driven organization, focusing on digital transformation and innovative insurance solutions. The lab aims to enhance operational capabilities, improve fraud detection, and support various insurance entities with advanced analytics. It emphasizes the importance of adapting technical solutions to local data environments while maintaining the ability to share insights across the organization.Xavier Conort, DataScience SG Meetup - Challenges in insurance pricing
Xavier Conort, DataScience SG Meetup - Challenges in insurance pricingKai Xin Thia
油
Modeling challenges for insurance pricing include:
1) Claims costs often follow skewed distributions with tails and are affected by discontinuities from policy limits, changing risk pools, and other factors.
2) Regulation and operational constraints require models to be simple and transparent while meeting requirements like premiums increasing with coverage.
3) Predicting future claims is difficult as models may not extrapolate to new situations and can be naive in how they learn from past data.How Lazada ranks products to improve customer experience and conversion
How Lazada ranks products to improve customer experience and conversionEugene Yan Ziyou
油
Lazada enhances product rankings to improve customer experience and conversion through data-driven methods that analyze user behavior and product engagement. The document discusses how Lazada employs machine learning and various metrics to identify and boost new and quality products, resulting in increased conversion rates and user engagement. Key methodologies involve measuring shopper interest, product similarity, and quality, alongside testing changes to ensure effective outcomes.Big data ppt
Big data pptNasrin Hussain
油
This document provides an overview of big data. It defines big data as large volumes of diverse data that are growing rapidly and require new techniques to capture, store, distribute, manage, and analyze. The key characteristics of big data are volume, velocity, and variety. Common sources of big data include sensors, mobile devices, social media, and business transactions. Tools like Hadoop and MapReduce are used to store and process big data across distributed systems. Applications of big data include smarter healthcare, traffic control, and personalized marketing. The future of big data is promising with the market expected to grow substantially in the coming years.Data science meetup - Spiros Antonatos
Data science meetup - Spiros AntonatosSpiros Antonatos
油
This document discusses the role of data scientists in large organizations and provides insights from the author's experience. It describes common daily activities like data cleaning, modeling, and reporting. Effective data scientists need skills in areas like statistics, machine learning, databases, and computer science. They must work closely with data engineers and domain experts to build the necessary data foundations and validate models. Visualization and understanding business needs are also important. Data scientists should learn multiple tools and technologies to adapt to different problems and work effectively in teams.Mclarens @ Data Science Sg
Mclarens @ Data Science SgBenji Thian
油
1) The document discusses how data and analytics can be used to change games in various industries like transportation, health and wellness, motorsports, energy, consumer, and rail.
2) It provides examples of how McLaren uses a data cycle of simulation, pre-race analysis, live monitoring and strategy, and post-race analysis with data management and visualization to improve performance.
3) The document outlines focus areas like human performance/condition monitoring, disease progression monitoring, and remaining useful life prediction using sensors, data capture, analytics, and other techniques to extract insights from data in areas like gait, biotelemetry, and health conditions.R language
R languageLearningTech
油
R is a free and open-source programming language and software environment for statistical analysis, graphics, and statistical computing. It was originally developed in the 1990s at Bell Laboratories by statisticians John Chambers and colleagues. Key points about R include that it is an interpreted language, supports functional programming, and is object-oriented. R can be used for tasks like statistical analysis, data visualization, and machine learning. It has a large community of users and developers contributing packages for specialized analysis techniques.Nielsen x DataScience SG Meetup (Apr 2015)
Nielsen x DataScience SG Meetup (Apr 2015)Eugene Yan Ziyou
油
The document discusses Nielsen's use of location-based analytics for marketing research. It describes Nielsen's global presence and services. Several case studies are presented, including using satellite imagery to classify rural vs urban areas in Indonesia for survey sampling. Nielsen applied machine learning techniques like k-means clustering to satellite images of Bali, Indonesia to develop an objective rural-urban classifier with high accuracy. This allows them to better stratify and manage large-scale survey fieldwork.Industrial internet big data uk market study
Industrial internet big data uk market studySari Ojala
油
The document is a 2015 research report on big data and the industrial internet in the UK. It finds that big data is a large and growing market opportunity in the UK, with the sector expected to deliver 贈6.8 billion in benefits to the economy by 2017. However, UK companies currently lag behind counterparts in Germany and France in adopting big data. The report identifies key industries using big data, such as financial services for algorithmic trading and fraud prevention. It also outlines opportunities for consulting, hosting, analytics software, and more.Competition Improves Performance: Only when Competition Form matches Goal Ori...
Competition Improves Performance: Only when Competition Form matches Goal Ori...Eugene Yan Ziyou
油
The document discusses how competition affects performance depending on an individual's goal orientation. It examines direct competition against others versus indirect competition against standards. Performance-oriented individuals may perform better in direct competition, while mastery-oriented individuals may perform better with indirect competition when the task is difficult. Two studies examined how goal orientation, perceived task difficulty or self-efficacy, and type of competition (direct vs indirect) interact to influence task performance. The first study found a trend that mastery-oriented individuals performed worse in direct competition when the task was difficult and perceived as difficult. The second study found marginal support that performance-oriented individuals performed better in direct competition when self-efficacy was high.Scalable Collaborative Filtering for Commerce Recommendation
Scalable Collaborative Filtering for Commerce RecommendationYiqun Hu
油
The document discusses a scalable solution for commerce recommendation using collaborative filtering and matrix factorization techniques. It compares the limitations of existing methods, such as Apache Mahout, to the proposed method, which significantly improves scalability with over 10x performance enhancement. The solution involves an efficient implementation using sharding and parallel processing to handle larger datasets effectively.Kaggle Otto Challenge: How we achieved 85th out of 3,514 and what we learnt
Kaggle Otto Challenge: How we achieved 85th out of 3,514 and what we learntEugene Yan Ziyou
油
The document outlines how the authors achieved the 85th position out of 3,514 teams in the Kaggle Otto product classification challenge, focused on minimizing multi-class log loss through various modeling techniques. They discuss data feature engineering, model validation methods, and the performance of different algorithms, including tree-based models and neural networks, while emphasizing the importance of ensemble methods. The authors also provide insights into strategies for Kaggle competitions, including effective validation processes and feature transformation techniques.Datalake Architecture
Datalake ArchitectureTechYugadi IT Solutions & Consulting
油
The document discusses data lake architecture, emphasizing its value in capturing and exploring vast amounts of raw data generated by businesses. It contrasts data lakes with enterprise data warehouses (EDW), outlining the importance of a scalable storage layer, flexible data processing, and robust governance practices. Additionally, it covers implementation steps and technology considerations such as using Hadoop for building a data lake.Big Data Paris - A Modern Enterprise Architecture
Big Data Paris - A Modern Enterprise ArchitectureMongoDB
油
The document discusses the evolving landscape of data management within enterprise architecture, emphasizing the challenges and opportunities posed by digital transformation. Key goals include enhancing operational intelligence, business agility, and customer-centricity, with solutions focusing on flexible data models and real-time analytics. The use of modern data architectures, like data-as-a-service and operationalized data lakes, is highlighted as essential for overcoming existing system limitations and improving customer experiences.Garuda Robotics x DataScience SG Meetup (Sep 2015)
Garuda Robotics x DataScience SG Meetup (Sep 2015)Eugene Yan Ziyou
油
The document discusses the use of drones in precision agriculture, specifically for palm oil plantations, highlighting their role in solving customer problems and gathering actionable data. It details workflows for data acquisition, analysis, and presentation, emphasizing the importance of predictive and prescriptive analytics over mere descriptive analysis. Case studies, such as tree counting, demonstrate the advanced capabilities of drone technology for enhancing agricultural efficiency and decision-making.Kudu: New Hadoop Storage for Fast Analytics on Fast Data
Kudu: New Hadoop Storage for Fast Analytics on Fast DataCloudera, Inc.
油
The document discusses Kudu, an open source storage system for Hadoop that is designed to enable both transactional and analytic workloads. Kudu uses a columnar storage format and provides ACID transactions for fast analytics on fast data. It aims to address gaps in Hadoop for workloads that require simultaneous random access and scanning of data. Benchmarks show Kudu can perform TPC-H queries within 2x of Parquet storage, with low latency for reads and writes on solid state drives.Featurizing log data before XGBoost
Featurizing log data before XGBoostDataRobot
油
The document discusses a project focused on predicting course dropout rates on a MOOC platform using XGBoost. It details the extensive feature engineering conducted by the team, including techniques like entropy calculations and sequence feature extraction from log data. Ultimately, the team successfully utilized these features to enhance model performance and achieve a notable increase in AUC scores.Deep Learning for Stock Prediction
Deep Learning for Stock PredictionLim Zhi Yuan (Zane)
油
This document describes research on using deep learning models to predict stock market movements based on news events. It presents a method to extract event representations from news articles, generalize the events, embed the events, and feed the embedded events into deep learning models. Experimental results show that using embedded events as inputs to convolutional neural networks achieved more accurate stock market predictions than baseline methods, and modeling long, mid, and short-term event effects further improved performance. The research demonstrates that deep learning can effectively capture hidden relationships between news events and stock prices.Death by PowerPoint
Death by PowerPointAlexei Kapterev
油
This document provides tips for creating effective PowerPoint presentations. It notes that many presentations are "unbearable" due to a lack of significance, structure, simplicity, and rehearsal. It emphasizes the importance of having a clear purpose for your presentation, using a simple structure like problem-solution, keeping slides concise with minimal text and images over clipart, writing speaker notes instead of long slides for printing, and rehearsing your presentation aloud to work out any issues. The overall message is that presentations should be passionate, memorable and scalable through a focus on simplicity and clarity of message.ICICCE0297
ICICCE0297IJTET Journal
油
This paper explores a data mining framework for identifying friends in online social networks when no link or content information is available. It highlights three social forceshomophily, confounding, and influencethat drive friendship formation and proposes methods to predict user attributes from usernames to facilitate friend discovery. The research suggests that while traditional methods struggle without initial social connections, utilizing minimal information can improve friend recommendation systems significantly.Multi-Mode Conceptual Clustering Algorithm Based Social Group Identification ...
Multi-Mode Conceptual Clustering Algorithm Based Social Group Identification ...inventionjournals
油
The document presents a multi-mode conceptual clustering algorithm aimed at enhancing collaborative web search by identifying social groups based on users' search contexts derived from web log data. It discusses the limitations of existing web search approaches and proposes a method that utilizes natural language processing for clustering users according to their interests, significantly reducing search time complexity. The proposed model was tested using real datasets and demonstrated superior accuracy and efficiency in interest prediction compared to traditional methods.Creating More Engaging Content For Social
Creating More Engaging Content For SocialEric T. Tung
油
This document summarizes Eric T. Tung's presentation on optimizing social media for engagement. It discusses using tools like Facebook Open Graph and Twitter Cards to generate rich previews when sharing content. It also recommends building Twitter lists to curate content and using automation tools to schedule posts to multiple networks. Metrics like retweets, likes and shares are discussed to measure social media impact.Jx2517481755
Jx2517481755IJERA Editor
油
This paper surveys link prediction and page ranking within blogs and social networks, emphasizing their significance in social media analytics. It discusses the methodologies used to gather and analyze blog data, including the dynamic interactions within blog communities and the predictions regarding social connections based on content sharing and engagement. Additionally, the findings highlight the integration of various analytical tools and models that aid in understanding social media's evolving landscape.Jx2517481755
Jx2517481755IJERA Editor
油
The document surveys link prediction and page ranking in blogs, highlighting social media's role in data mining and analytics. It discusses various tools and methods for analyzing social networks and blogs, including the significance of hyperlinks and the importance of sentiment analysis and web mining. Additionally, it explores the implications of blogging for communication and community engagement while addressing the challenges of analyzing dynamic content and predicting user behavior.Hubspot social-media-secrets-
Hubspot social-media-secrets-Say Digital Media
油
The document outlines 20 lesser-known features across Facebook, Twitter, LinkedIn, and Pinterest that enhance user experience and functionality. These features include saving links for later on Facebook, embedding presentations in tweets, adding media to LinkedIn profiles, and using trackable links on Pinterest. The guide serves as a resource for users looking to leverage these social media platforms more effectively.20 Facebook, Twitter, Linkedin & Pinterest Features You Didn't Know Existed (...
20 Facebook, Twitter, Linkedin & Pinterest Features You Didn't Know Existed (...HubSpot
油
The document outlines 20 hidden or lesser-known features of popular social media platforms like Facebook, Twitter, LinkedIn, and Pinterest. Some of the featured tips include saving links and articles to access later on Facebook, replacing ads with baby animal pictures on Facebook using a Chrome extension, embedding 際際滷Share presentations directly into tweets, and creating a photo collage directly in a single tweet on Twitter. The document also provides tips for adding hidden relationship notes on LinkedIn profiles and using trackable links on Pinterest that won't get flagged as spam.Hubspot Social Media Secrets
Hubspot Social Media SecretsJoaqu鱈n Moral P辿rez
油
The document highlights 20 lesser-known features across Facebook, Twitter, LinkedIn, and Pinterest that enhance user experience and engagement on these social networks. It covers various functionalities including saving links, embedding media, polling, and tracking interactions with content. These features aim to maximize the utility of social media platforms for both personal and professional use.20 Facebook, Twitter, LinkedIn & Pinterest Features You Didn't Know Existed B...
20 Facebook, Twitter, LinkedIn & Pinterest Features You Didn't Know Existed B...Black Marketing
油
The document describes 20 hidden or lesser-known features of Facebook, Twitter, LinkedIn, and Pinterest. Some of the featured social media tools include saving links and articles on Facebook for later, replacing Facebook ads with pictures of pets using a Chrome extension, embedding 際際滷Share presentations in tweets, analyzing tweet performance with Twitter Analytics, tagging people in photos on Twitter, and creating a photo collage with up to 4 photos in a single tweet on Twitter. The document is aimed at helping users discover new ways to use popular social networks.Hubspot social media secrets
Hubspot social media secretsJamie Christensen
油
The document describes 20 hidden or lesser-known features of Facebook, Twitter, LinkedIn, and Pinterest. Some of the features covered include saving links on Facebook to read later, replacing Facebook ads with pictures of pets using a Chrome extension, embedding 際際滷Share presentations directly into tweets, adding hidden relationship notes on LinkedIn profiles, and saving job searches on LinkedIn. The document is intended to help users discover these overlooked features on major social networks.More Related Content
Viewers also liked (12)
Industrial internet big data uk market study
Industrial internet big data uk market studySari Ojala
油
The document is a 2015 research report on big data and the industrial internet in the UK. It finds that big data is a large and growing market opportunity in the UK, with the sector expected to deliver 贈6.8 billion in benefits to the economy by 2017. However, UK companies currently lag behind counterparts in Germany and France in adopting big data. The report identifies key industries using big data, such as financial services for algorithmic trading and fraud prevention. It also outlines opportunities for consulting, hosting, analytics software, and more.Competition Improves Performance: Only when Competition Form matches Goal Ori...
Competition Improves Performance: Only when Competition Form matches Goal Ori...Eugene Yan Ziyou
油
The document discusses how competition affects performance depending on an individual's goal orientation. It examines direct competition against others versus indirect competition against standards. Performance-oriented individuals may perform better in direct competition, while mastery-oriented individuals may perform better with indirect competition when the task is difficult. Two studies examined how goal orientation, perceived task difficulty or self-efficacy, and type of competition (direct vs indirect) interact to influence task performance. The first study found a trend that mastery-oriented individuals performed worse in direct competition when the task was difficult and perceived as difficult. The second study found marginal support that performance-oriented individuals performed better in direct competition when self-efficacy was high.Scalable Collaborative Filtering for Commerce Recommendation
Scalable Collaborative Filtering for Commerce RecommendationYiqun Hu
油
The document discusses a scalable solution for commerce recommendation using collaborative filtering and matrix factorization techniques. It compares the limitations of existing methods, such as Apache Mahout, to the proposed method, which significantly improves scalability with over 10x performance enhancement. The solution involves an efficient implementation using sharding and parallel processing to handle larger datasets effectively.Kaggle Otto Challenge: How we achieved 85th out of 3,514 and what we learnt
Kaggle Otto Challenge: How we achieved 85th out of 3,514 and what we learntEugene Yan Ziyou
油
The document outlines how the authors achieved the 85th position out of 3,514 teams in the Kaggle Otto product classification challenge, focused on minimizing multi-class log loss through various modeling techniques. They discuss data feature engineering, model validation methods, and the performance of different algorithms, including tree-based models and neural networks, while emphasizing the importance of ensemble methods. The authors also provide insights into strategies for Kaggle competitions, including effective validation processes and feature transformation techniques.Datalake Architecture
Datalake ArchitectureTechYugadi IT Solutions & Consulting
油
The document discusses data lake architecture, emphasizing its value in capturing and exploring vast amounts of raw data generated by businesses. It contrasts data lakes with enterprise data warehouses (EDW), outlining the importance of a scalable storage layer, flexible data processing, and robust governance practices. Additionally, it covers implementation steps and technology considerations such as using Hadoop for building a data lake.Big Data Paris - A Modern Enterprise Architecture
Big Data Paris - A Modern Enterprise ArchitectureMongoDB
油
The document discusses the evolving landscape of data management within enterprise architecture, emphasizing the challenges and opportunities posed by digital transformation. Key goals include enhancing operational intelligence, business agility, and customer-centricity, with solutions focusing on flexible data models and real-time analytics. The use of modern data architectures, like data-as-a-service and operationalized data lakes, is highlighted as essential for overcoming existing system limitations and improving customer experiences.Garuda Robotics x DataScience SG Meetup (Sep 2015)
Garuda Robotics x DataScience SG Meetup (Sep 2015)Eugene Yan Ziyou
油
The document discusses the use of drones in precision agriculture, specifically for palm oil plantations, highlighting their role in solving customer problems and gathering actionable data. It details workflows for data acquisition, analysis, and presentation, emphasizing the importance of predictive and prescriptive analytics over mere descriptive analysis. Case studies, such as tree counting, demonstrate the advanced capabilities of drone technology for enhancing agricultural efficiency and decision-making.Kudu: New Hadoop Storage for Fast Analytics on Fast Data
Kudu: New Hadoop Storage for Fast Analytics on Fast DataCloudera, Inc.
油
The document discusses Kudu, an open source storage system for Hadoop that is designed to enable both transactional and analytic workloads. Kudu uses a columnar storage format and provides ACID transactions for fast analytics on fast data. It aims to address gaps in Hadoop for workloads that require simultaneous random access and scanning of data. Benchmarks show Kudu can perform TPC-H queries within 2x of Parquet storage, with low latency for reads and writes on solid state drives.Featurizing log data before XGBoost
Featurizing log data before XGBoostDataRobot
油
The document discusses a project focused on predicting course dropout rates on a MOOC platform using XGBoost. It details the extensive feature engineering conducted by the team, including techniques like entropy calculations and sequence feature extraction from log data. Ultimately, the team successfully utilized these features to enhance model performance and achieve a notable increase in AUC scores.Deep Learning for Stock Prediction
Deep Learning for Stock PredictionLim Zhi Yuan (Zane)
油
This document describes research on using deep learning models to predict stock market movements based on news events. It presents a method to extract event representations from news articles, generalize the events, embed the events, and feed the embedded events into deep learning models. Experimental results show that using embedded events as inputs to convolutional neural networks achieved more accurate stock market predictions than baseline methods, and modeling long, mid, and short-term event effects further improved performance. The research demonstrates that deep learning can effectively capture hidden relationships between news events and stock prices.Death by PowerPoint
Death by PowerPointAlexei Kapterev
油
This document provides tips for creating effective PowerPoint presentations. It notes that many presentations are "unbearable" due to a lack of significance, structure, simplicity, and rehearsal. It emphasizes the importance of having a clear purpose for your presentation, using a simple structure like problem-solution, keeping slides concise with minimal text and images over clipart, writing speaker notes instead of long slides for printing, and rehearsing your presentation aloud to work out any issues. The overall message is that presentations should be passionate, memorable and scalable through a focus on simplicity and clarity of message.Similar to Social network analysis and growth recommendations for DataScience SG community (10)
ICICCE0297
ICICCE0297IJTET Journal
油
This paper explores a data mining framework for identifying friends in online social networks when no link or content information is available. It highlights three social forceshomophily, confounding, and influencethat drive friendship formation and proposes methods to predict user attributes from usernames to facilitate friend discovery. The research suggests that while traditional methods struggle without initial social connections, utilizing minimal information can improve friend recommendation systems significantly.Multi-Mode Conceptual Clustering Algorithm Based Social Group Identification ...
Multi-Mode Conceptual Clustering Algorithm Based Social Group Identification ...inventionjournals
油
The document presents a multi-mode conceptual clustering algorithm aimed at enhancing collaborative web search by identifying social groups based on users' search contexts derived from web log data. It discusses the limitations of existing web search approaches and proposes a method that utilizes natural language processing for clustering users according to their interests, significantly reducing search time complexity. The proposed model was tested using real datasets and demonstrated superior accuracy and efficiency in interest prediction compared to traditional methods.Creating More Engaging Content For Social
Creating More Engaging Content For SocialEric T. Tung
油
This document summarizes Eric T. Tung's presentation on optimizing social media for engagement. It discusses using tools like Facebook Open Graph and Twitter Cards to generate rich previews when sharing content. It also recommends building Twitter lists to curate content and using automation tools to schedule posts to multiple networks. Metrics like retweets, likes and shares are discussed to measure social media impact.Jx2517481755
Jx2517481755IJERA Editor
油
This paper surveys link prediction and page ranking within blogs and social networks, emphasizing their significance in social media analytics. It discusses the methodologies used to gather and analyze blog data, including the dynamic interactions within blog communities and the predictions regarding social connections based on content sharing and engagement. Additionally, the findings highlight the integration of various analytical tools and models that aid in understanding social media's evolving landscape.Jx2517481755
Jx2517481755IJERA Editor
油
The document surveys link prediction and page ranking in blogs, highlighting social media's role in data mining and analytics. It discusses various tools and methods for analyzing social networks and blogs, including the significance of hyperlinks and the importance of sentiment analysis and web mining. Additionally, it explores the implications of blogging for communication and community engagement while addressing the challenges of analyzing dynamic content and predicting user behavior.Hubspot social-media-secrets-
Hubspot social-media-secrets-Say Digital Media
油
The document outlines 20 lesser-known features across Facebook, Twitter, LinkedIn, and Pinterest that enhance user experience and functionality. These features include saving links for later on Facebook, embedding presentations in tweets, adding media to LinkedIn profiles, and using trackable links on Pinterest. The guide serves as a resource for users looking to leverage these social media platforms more effectively.20 Facebook, Twitter, Linkedin & Pinterest Features You Didn't Know Existed (...
20 Facebook, Twitter, Linkedin & Pinterest Features You Didn't Know Existed (...HubSpot
油
The document outlines 20 hidden or lesser-known features of popular social media platforms like Facebook, Twitter, LinkedIn, and Pinterest. Some of the featured tips include saving links and articles to access later on Facebook, replacing ads with baby animal pictures on Facebook using a Chrome extension, embedding 際際滷Share presentations directly into tweets, and creating a photo collage directly in a single tweet on Twitter. The document also provides tips for adding hidden relationship notes on LinkedIn profiles and using trackable links on Pinterest that won't get flagged as spam.Hubspot Social Media Secrets
Hubspot Social Media SecretsJoaqu鱈n Moral P辿rez
油
The document highlights 20 lesser-known features across Facebook, Twitter, LinkedIn, and Pinterest that enhance user experience and engagement on these social networks. It covers various functionalities including saving links, embedding media, polling, and tracking interactions with content. These features aim to maximize the utility of social media platforms for both personal and professional use.20 Facebook, Twitter, LinkedIn & Pinterest Features You Didn't Know Existed B...
20 Facebook, Twitter, LinkedIn & Pinterest Features You Didn't Know Existed B...Black Marketing
油
The document describes 20 hidden or lesser-known features of Facebook, Twitter, LinkedIn, and Pinterest. Some of the featured social media tools include saving links and articles on Facebook for later, replacing Facebook ads with pictures of pets using a Chrome extension, embedding 際際滷Share presentations in tweets, analyzing tweet performance with Twitter Analytics, tagging people in photos on Twitter, and creating a photo collage with up to 4 photos in a single tweet on Twitter. The document is aimed at helping users discover new ways to use popular social networks.Hubspot social media secrets
Hubspot social media secretsJamie Christensen
油
The document describes 20 hidden or lesser-known features of Facebook, Twitter, LinkedIn, and Pinterest. Some of the features covered include saving links on Facebook to read later, replacing Facebook ads with pictures of pets using a Chrome extension, embedding 際際滷Share presentations directly into tweets, adding hidden relationship notes on LinkedIn profiles, and saving job searches on LinkedIn. The document is intended to help users discover these overlooked features on major social networks.Ad
More from Eugene Yan Ziyou (15)
System design for recommendations and search
System design for recommendations and searchEugene Yan Ziyou
油
The document summarizes Eugene Yan's presentation on system design for recommendations and search. It discusses batch vs real-time systems, the offline and online environments, candidate retrieval vs ranking models, and examples from Alibaba, Facebook, and DoorDash. It recommends building an MVP by using self-supervised learning for embeddings, approximate nearest neighbors for retrieval, logistic regression for ranking, and serving the system with load balancing across instances.Recommender Systems: Beyond the user-item matrix
Recommender Systems: Beyond the user-item matrixEugene Yan Ziyou
油
The document discusses the implementation and results of recommender systems beyond the traditional user-item matrix, with a focus on data preparation, parsing JSON to CSV, and creating product pairs. It details various techniques for matrix factorization, including training schedules, scoring relationships, and generating negative samples. The document also explores integrating natural language processing and graph methods to enhance recommendation accuracy.Predicting Hospital Bills at Pre-admission
Predicting Hospital Bills at Pre-admissionEugene Yan Ziyou
油
Ucare.ai utilizes machine learning and data engineering to predict healthcare costs, aiming to enhance patient experience and minimize payment challenges. The system addresses deficiencies in existing estimation methods by providing timely and accurate cost predictions, reducing underestimation incidents significantly. Its integration maintains minimal disruption to frontline users while enabling frequent updates and improved user experiences.OLX Group Prod Tech 2019 Keynote: Asia's Tech Giants
OLX Group Prod Tech 2019 Keynote: Asia's Tech GiantsEugene Yan Ziyou
油
The document discusses the scaling strategies of Asia's tech giants, focusing on Alibaba and Grab, and their respective approaches to creating superapps. It contrasts Alibaba's centralized capabilities with Grab's hyper-localization strategy, highlighting the importance of data synergies and local context in platform development. The document also outlines the evolution of superapps in China and Southeast Asia, examining the reasons for their success and the unbundling trend occurring in these markets.Data Science Challenges and Impact at Lazada (Big Data and Analytics Innovati...
Data Science Challenges and Impact at Lazada (Big Data and Analytics Innovati...Eugene Yan Ziyou
油
The document discusses data science challenges faced by Lazada, focusing on the trade-offs between business input and automated algorithms in product ranking and development speed versus production stability. It highlights the importance of collaboration between business needs and data innovation, exemplified by timebox-ed skunkworks resulting in proofs of concept. Key takeaways emphasize that effective data science involves a mix of coding, machine learning, and various activities, depending on team maturity and organizational needs.INSEAD Sharing on Lazada Data Science and my Journey
INSEAD Sharing on Lazada Data Science and my JourneyEugene Yan Ziyou
油
Lazada, a leading e-commerce platform in Southeast Asia, leverages data science to enhance its operations, including user tracking, marketing, and fraud management. The document outlines the author's journey in data science, highlighting their experience in various roles and key data projects focused on improving product ranking, consumer recommendations, and operational efficiencies. It concludes with suggestions for building data analytics skills and emphasizes the importance of effective communication and practical experience.SMU BIA Sharing on Data Science
SMU BIA Sharing on Data ScienceEugene Yan Ziyou
油
The document summarizes an information session on data analytics from SMU Business Intelligence and Analytics. It defines data analytics as using analysis and visualization to provide insights and solve problems. It also discusses using machine learning to automate problem solving. The document then provides recommendations on learning tools like SQL, Python, R, and Spark through online courses. It suggests gaining skills in areas like statistics, machine learning, and communication. Finally, it shares one person's journey entering data analytics through studying psychology and business, then working on data analysis projects at IBM and Lazada.Culture at Lazada Data Science
Culture at Lazada Data ScienceEugene Yan Ziyou
油
The document outlines the data science culture at Lazada, emphasizing the importance of shared beliefs, values, and practices that foster professional growth and a collaborative workplace. It highlights core values such as ownership, communication, innovation, and impact, while advocating for responsible decision-making and hiring practices that seek to elevate team talent. Additionally, it promotes a culture of learning from failures and encourages proactive engagement in one's professional development.Sharing about my data science journey and what I do at Lazada
Sharing about my data science journey and what I do at LazadaEugene Yan Ziyou
油
Eugene shares his data science journey and experience at Lazada, detailing his educational background and professional roles, including work at IBM and the Ministry of Trade & Industry. He discusses the skills acquired through education and various MOOCs, as well as practical experience through volunteering and Kaggle competitions. Currently, at Lazada, he focuses on product categorization, recommendations, and consumer segmentation, while emphasizing the importance of coding, data preparation, and stakeholder engagement in his work.DataKind SG sharing of our first DataDive
DataKind SG sharing of our first DataDiveEugene Yan Ziyou
油
The document details a post-Datadive sharing event hosted by DataKind Singapore on July 23, 2015, covering updates, an overview of Datadive events, and data handling best practices. It highlights various projects, including a hackathon promoting financial literacy and analysis of Earth Hour tweets for sentiment and influencer identification. The document emphasizes the importance of collaboration and structured data handling to maximize the impact of data-driven initiatives for social good.Statistical inference: Statistical Power, ANOVA, and Post Hoc tests
Statistical inference: Statistical Power, ANOVA, and Post Hoc testsEugene Yan Ziyou
油
This document provides an overview of statistical power, analysis of variance (ANOVA), and post hoc tests. It defines statistical power and explains how to calculate power and minimum sample size. It then describes ANOVA, comparing it to t-tests. ANOVA partitions variability between and within groups. The document interprets ANOVA tables and explains F distributions. Conditions for ANOVA and post hoc tests like Bonferroni corrections are also covered. Finally, it briefly mentions different types of ANOVA like one-way and factorial.Statistical inference: Hypothesis Testing and t-tests
Statistical inference: Hypothesis Testing and t-testsEugene Yan Ziyou
油
This document covers statistical inference concepts, focusing on hypothesis testing and t-tests in relation to primary school children's height in Singapore. It explains how to calculate confidence intervals, sample size for desired margins of error, and the implications of p-values in hypothesis testing. The document also discusses the use of t-distribution versus normal distribution, dependent and independent t-tests, and the importance of effect size in evaluating practical significance.Statistical inference: Probability and Distribution
Statistical inference: Probability and DistributionEugene Yan Ziyou
油
The document covers statistical inference, focusing on probability concepts, types of variables, and various important distributions such as Bernoulli, binomial, normal, and Poisson. It explains essential probability rules, joint and conditional probabilities, and Bayes' theorem, alongside methods for calculating expected values and variances. Additionally, it provides examples and applications of these concepts in real-world contexts, such as cancer detection and ad campaign performance.A Study on the Relationship between Education and Income in the US
A Study on the Relationship between Education and Income in the USEugene Yan Ziyou
油
The document examines the relationship between education level and income using data from the 2012 General Social Survey. It finds a significant positive correlation between higher education levels and income, with notable conditions indicating that graduates earn substantially more than those with lower education, although there is no significant difference in income between high school and junior college graduates. The study acknowledges limitations including the use of total family income instead of personal income and the lack of inclusivity for non-English or Spanish speakers.Diving into Twitter data on consumer electronic brands
Diving into Twitter data on consumer electronic brandsEugene Yan Ziyou
油
This document summarizes an analysis of Twitter data related to consumer electronic brands collected between May 13-25, 2014. It was found that Apple and iPhone accounted for 87% of tweet volume. The data was optimized by converting 15.3GB of JSON data to 757MB of CSV format. Tweets mentioning seven brands were analyzed for sentiment and found to have roughly equal ratios of positive to negative tweets, except Samsung which had a ratio of roughly 3:2. Spikes in tweet volume coincided with product launches. The analysis revealed differences in users tweeting about different brands as well as opportunities for future analysis.Ad
Social network analysis and growth recommendations for DataScience SG community
- 6. Smaller nodes represent long inactive periods Colored from blue (inactive for more than 3 months) to red (active in last one month)
- 9. with locked positions from friendship network Yellow edges- liking posts from non-friends Red edges - liking posts from friends
- 11. User-User like network Colored via Markov Clustering Nodes sized by number of likes given
- 13. User-user like network Examples of one-off likes who did not like any other posts.
- 17. User-user comment connection, using frozen positions from like network previously. Colored by Markov Clustering from the like network
- 18. User-user comment network. Colored by Markov Clustering from the like network
- 22. - - - - -
- 24. Use substrings to extract full web links Use Python scripts to unmask the redirect link
- 25. Copy the code above to macro The new function called Title will be generated Drag down as usual
- 27. Remove punctuation Remove Numbers Remove stopwords such as analyst,data,singapore and the website name such as straits time Perform LDA get top 4 topics and top 7 most frequency words in the topic