Let’s look at examples of machine learning for each of the five problems to be solved.
Taxi Dispatch Prediction (Transportation)
Prediction of passengers waiting for taxis, supplemented by local familiarity and experience that veteran drivers had until now, is now possible with high accuracy even for new drivers regardless of knowledge by using a prediction model using AI technology.
It is possible to move to such a place, improve the taxi occupancy rate, and increase sales. It is also expected to enhance user satisfaction by reducing user waiting times.
Production Forecast/Growth Forecast (Agriculture)
When shipping crops, we will share prediction results such as harvest volume and harvest time obtained using growth data, greenhouse environment data, and weather data and use them in the production process of crops to improve quality and yield—securing the necessary personnel and distribution destinations in advance for harvesting work, maintaining stable harvest volumes and increasing sales prices by being able to trade under favourable conditions.
It can be expected. It is also likely to be effective in predicting crops’ price and demand and reducing crop waste.
Demand Forecast (Apparel/retail)
In the apparel industry, where trends change quickly, and excess inventory and increased waste have become problems, we can predict trends from distribution data, etc., and use them in product planning and inventory management to improve sales and customer purchasing behaviour. It is expected to have the effect of reducing excess inventory.
In addition to the apparel industry, it is also used in retail, vending machines and convenience stores. AI is also used to order products based on demand forecasts. Many major supermarkets and convenience stores are working on this, as productivity will increase by putting the right people in the right place, such as shortening the ordering time and allowing employees to focus on customer service. Increase.
Inquiry Response By Chatbot (Office Work)
Some chatbots facilitate communication with customers, such as responding to inquiries from the web, and those that facilitate communication with employees, such as responding to internal help desks. By enabling both to be performed automatically, we can expect to reduce the burden on customer support, shorten the time required for inquiries, and reduce human costs. The number of companies that are introducing it is increasing.
Call Centre Automation (Business)
While the issue of securing human resources for call centres is becoming more serious, call centres have traditionally been operated entirely by human power, so they can only handle a limited number of people during business hours, making it difficult to connect due to the concentration of calls. There were issues, such as wanting to be able to respond outside of the daytime.
By making it possible to answer calls automatically with voice recognition, it is now possible to respond outside business hours, and it has become possible to respond to more customers. It is expected to reduce human costs at call centres, improve user satisfaction and convenience, and is being considered for introduction in various industries.
Digitization Of Handwritten Documents And Automation Of Input (Office Work)
Substituting a system that recognizes handwritten characters and automatically inputs documents, which used to be done manually, dramatically reduces the work time. It is being introduced in various industries, such as logistics, insurance, and finance.
Stock Price Prediction (Finance)
In stock Investment, it was necessary to be familiar with information such as financial results and stock price data and to consult with financial experts by paying expensive fees. If prediction becomes widespread, it may become easier to start. In addition, it will be possible to buy and sell at the right time, which can be expected to reduce transaction costs.
Automatic Sentence Summarization (Media)
It took time and effort to summarize sentences such as news articles, and it was necessary to hire skilled workers and train workers to maintain a consistent level of summarization, so securing personnel was an issue.
You can expect to reduce costs by summarising sentences automatically while maintaining uniform quality.
Automatic Generation Of Digest Videos (Media)
In the past, highlight scenes and digest videos of matches were created manually. Still, AI can automatically generate videos focusing on specific players and extract scenes with highlights and characteristics. In addition to shortening production time and reducing labour costs; users can also expect to improve satisfaction, such as being able to check the highlights immediately after the game.
Road Surface Deterioration Diagnosis (Detection Of Cracks In Concrete) (Infrastructure)
As with the inspection of structures above, the roads built during high economic growth in the 1960s and 1970s are ageing. There are many activities, such as demonstration experiments, to automate the judgement of road damage.
It can detect even minute cracks that are difficult to see with the human eye and can be expected to speed up repair work and reduce work time and costs.
Automatic Ad Creative Generation (Advertisement/Creative)
To solve the problem of shortening the production time of advertising creatives, tools are being developed and researched to automatically generate many advertising proposals that are predicted to be highly effective. You can expect cost reduction by shortening the production time and improvement in the diversity of candidate proposals.
Editing And Proofreading Support (Publishing)
For printed materials such as pamphlets and advertising productions, specific staff members check and proofread the content of the manuscripts, and the burden of checking on the confirming side and proofreading skills become dependent on the individual. Therefore, AI can learn notation and technical terms specific to an industry or company, detect typographical errors according to standards, and make it possible to review and proofread sentences, reducing the burden on workers and reducing human error. is realized.
Diagnosis Support By Cancer Detection (Medical)
In the past, there was a risk of doctors overlooking lesions during endoscopy, and there were concerns about the burden on doctors due to the large number of double checks performed visually in a limited amount of time. Therefore, efforts have begun to use AI to detect lesions from endoscopic images, prevent cancer from being overlooked, and reduce the burden on doctors.
Nutritional Management By Calorie Estimation Of Meals (Food)
People often try to record their meals for reasons such as diet and prevention of lifestyle-related diseases, but keeping track of their meals is difficult. I was. In addition to the general public, the trial introduction is progressing for the diet management of athletes.
Fraudulent Transaction/Remittance Detection (Finance)
By streamlining the stock transaction screening work conventionally done manually by screening personnel to prevent fraudulent transactions, the screening personnel will be able to investigate and analyse more complex and sophisticated fraudulent methods. It is expected to improve detection accuracy.
Driver Safety Management (Automobile)
The number of accidents caused by driving while using a smartphone is increasing; the development and introduction of AI-equipped drive recorders and AI cameras installed on the road are progressing for driver safety management, such as preventing falling asleep while driving and driving.
Automating Traffic Volume Surveys (Traffic)
Traffic volume surveys, traditionally done visually by humans, can be conducted over a long period at a low cost by automatically measuring the number of vehicles and pedestrians using AI-based vehicle and pedestrian detection and tracking technology.
It is now possible. Action screening works conventionally done manually by screening personnel to prevent fraudulent transactions; the screening personnel will be able to investigate and analyse more complex and sophisticated fraudulent methods. It is expected to improve detection accuracy.
Parked Vehicle Estimate (Retail)
It is now possible to analyse the number of visitors by counting the number of cars parked in the store’s parking lot from satellite data. You can also predict the number of visitors by conducting fixed-point observations.
As a result, it will be possible to grasp not only the situation of the company but also other companies, so it is likely to lead to analysis such as where to open a store.