AU2023200945A1 - Method and Apparatus for Asset Selection - Google Patents

Abstract

Disclosed is a method of selecting an asset for investment. The asset selection method includes receiving a plurality of historical features related to an asset type of the asset, the historical features including information related to the asset type, and generating a machine learning model for the asset type using the plurality of historical features. The asset selection method also includes estimating a long term relative growth for an asset by applying the features related to the asset to the machine learning model and comparing the estimated long term relative growth to a predetermined threshold to select the asset as an investment. 1/7 100 Input data 118 Processor Interface 112 Input 106 102 device 110 Bus 104 Memory Output 108 116 device Database Storage device 120 114 Output data Fig. 1

Description

1/7

100

Input data

118

Interface 112 Input 106 Processor 102 device

110

Bus

104 Memory Output 108 116 device Database

Storage device 120

114 Output data

Fig. 1

Australian Patents Act 1990

ORIGINAL COMPLETE SPECIFICATION STANDARDPATENT

Invention Title Method and Apparatus for Asset Selection

The following statement is a full description of this invention, including the best method of performing it known to me/us:

- la

Technical Field

[001] The present invention relates to asset selection, and particularly to a method and apparatus for selecting an asset for long term investment.

Background

[002] Real estate co-investment enables a property buyer an opportunity to purchase a real estate asset in conjunction with another party. Such an arrangement may allow the purchaser an opportunity to purchase property they otherwise could not. The property buyer can be a first time homeowner with insufficient savings for a down payment, a current homeowner wishing to purchase a new home, or an investor looking to invest in a property acquisition.

[003] One issue with co-investment is that prospective buyers want full control and ownership of their asset. That is, the prospective buyer typically wants 100% ownership. Such an arrangement may be attractive because it will be appealing to prospective buyers. However, giving 100% ownership to the prospective buyer results in a loss of control for any co-investor.

[004] To compensate for the risk arising from the loss of control for the asset by the co investor, a higher return may be desired. If the co-investor takes too high a share of proceeds on eventual profits, when the asset is sold, the co-investment arrangement may no longer be desirable for the potential buyer.

[005] Typically, selection of suitable assets, such as a real estate asset, is made using the advice of professional with skills relating to the asset. The professional may use market data and experience to determine the suitability of an asset for investment. However, the skills of such a professional may vary between individuals. Further, their ability to select suitable assets and provide guidance for the return of the asset may be limited to certain geographical areas with knowledge of one area not being applicable to another area.

Summary

[006] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.

[007] One embodiment includes a method of selecting an asset for investment, the method comprising: receiving a plurality of historical features related to an asset type of the asset, the historical features including information related to the asset type; generating a machine learning model for the asset type using the plurality of historical features; estimating a long term relative growth for an asset by applying the features related to the asset to the machine learning model; and comparing the estimated long term relative growth to a predetermined threshold to select the asset as an investment.

[008] In one embodiment the long term relative growth is a predicted percentile value of growth for the asset.

[009] In one embodiment the predicted percentile value for the asset is for the asset after a fixed time period.

[010] In one embodiment the predetermined threshold is determined using historical data.

[011] In one embodiment the plurality of historical features related to the asset type comprise at least one data type selected from a list of data types consisting of asset data, market data, and high level data.

[012] In one embodiment the predetermined machine learning model is a gradient boosted machine.

[013] In one embodiment the gradient boosted machine is an extreme gradient boosted machine.

[014] In one embodiment the machine learning model is generated based on a fixed time period for the investment term.

[015] One embodiment includes an apparatus for selecting an asset comprising a processor, wherein the apparatus is configured to: receive a plurality of historical features related to an asset type of the asset, the historical features including information related to the asset type; generate a machine learning model for the asset type using the plurality of historical features; estimate a long term relative growth for an asset by applying the features related to the asset to the machine learning model; and compare the estimated long term relative growth to a predetermined threshold to select the asset as an investment.

[016] A method for generating a machine learning model for estimating long term relative growth of an asset type, the method comprising: receiving a plurality of historical features related to the asset type, the historical features including information related to the asset type and being selected based on a fixed time period for an investment term; and generating a machine learning model for the asset type of the asset and the fixed time period.

Brief Description of Figures

[017] At least one embodiment of the present invention is described, by way of example only, with reference to the accompanying figures.

[018] Figure 1 illustrates a functional block diagram of an example processing system that can be utilised to embody or give effect to a particular embodiment;

[019] Figure 2 illustrates an example network infrastructure that can be utilised to embody or give effect to a particular embodiment;

[020] Figures 3 illustrates a system generation process according to one embodiment;

[021] Figure 4 illustrates a feature table used in the system generation process of Figure 3;

[022] Figure 5 illustrates an asset evaluation process according to one embodiment;

[023] Figure 6 illustrates an asset evaluation process according to one embodiment; and

[024] Figures 7A and 7B illustrate a training system and an asset assessor according to one embodiment.

Detailed Description

[025] The following modes, given by way of example only, are described in order to provide a more precise understanding of one or more embodiments. In the figures, like reference numerals are used to identify like parts throughout the figures.

[026] Disclosed is a novel solution that may avoid a co-investor taking too high a share of proceeds by selecting assets that may be more likely to outperform the market over the long term. Such an approach may compensate a co-investor with a return that may compensate for any loss of control of the asset.

[027] Disclosed is a method of selecting an asset for investment. The method includes receiving a plurality of historical features related to an asset type of the asset, the historical features including information related to the asset type. The method generates a machine learning model for the asset type using the plurality of historical features. The method estimates a long term relative growth for an asset by applying the features related to the asset to the machine learning model and compares the estimated long term relative growth to a predetermined threshold to select the asset as an investment.

[028] A particular embodiment of the present invention can be realised using a processing system, an example of which is shown in Fig. 1. In particular, the processing system 100 generally includes at least one processor 102, or processing unit or plurality of processors, memory 104, at least one input device 106 and at least one output device 108, coupled together via a bus or group of buses 110. In certain embodiments, input device 106 and output device 108 could be the same device. An interface 112 can also be provided for coupling the processing system 100 to one or more peripheral devices, for example interface 112 could be a PCI card or PC card. At least one storage device 114 which houses at least one database 116 can also be provided. The memory 104 can be any form of memory device, for example, volatile or non-volatile memory, solid state storage devices, magnetic devices, etc. The processor 102 could include more than one distinct processing device, for example to handle different functions within the processing system 100.

[029] Input device 106 receives input data 118 and can include, for example, a keyboard, a pointer device such as a pen-like device or a mouse, audio receiving device for voice controlled activation such as a microphone, data receiver or antenna such as a modem or wireless data adaptor, data acquisition card, etc. Input data 118 could come from different sources, for example keyboard instructions in conjunction with data received via a network. Output device 108 produces or generates output data 120 and can include, for example, a display device or monitor in which case output data 120 is visual, a printer in which case output data 120 is printed, a port for example a USB port, a peripheral component adaptor, a data transmitter or antenna such as a modem or wireless network adaptor, etc. Output data 120 could be distinct and derived from different output devices, for example a visual display on a monitor in conjunction with data transmitted to a network. A user could view data output, or an interpretation of the data output, on, for example, a monitor or using a printer. The storage device 114 can be any form of data or information storage means, for example, volatile or non-volatile memory, solid state storage devices, magnetic devices, etc.

[030] In use, the processing system 100 is adapted to allow data or information to be stored in and/or retrieved from, via wired or wireless communication means, the at least one database 116. The interface 112 may allow wired and/or wireless communication between the processing unit 102 and peripheral components that may serve a specialised purpose. The processor 102 receives instructions as input data 118 via input device 106 and can display processed results or other output to a user by utilising output device 108. More than one input device 106 and/or output device 108 can be provided. It should be appreciated that the processing system 100 may be any form of terminal, server, specialised hardware, or the like. In some instances, such as when a touch screen display is used, the output device 108 and the input device 106 may be the same device.

[031] The processing system 100 may be a part of a networked communications system 200, as shown in Fig. 2. Processing system 100 could connect to network 202, for example the Internet or a WAN. Input data 118 and output data 120 could be communicated to other devices via network 202. Other terminals, for example, thin client 204, further processing systems 206 and 208, notebook computer 210, mainframe computer 212, PDA 214, pen based computer or tablet 216, server 218, etc., can be connected to network 202. A large variety of other types of terminals or configurations could be utilised. The transfer of information and/or data over network 202 can be achieved using wired communications means 220 or wireless communications means 222. Server 218 can facilitate the transfer of data between network 202 and one or more databases 224. Server 218 and one or more databases 224 provide an example of an information source.

[032] Other networks may communicate with network 202. For example, telecommunications network 230 could facilitate the transfer of data between network 202 and mobile, cellular telephone or smartphone 232 or a PDA-type device 234, by utilising wireless communication means 236 and receiving/transmitting station 238. Satellite communications network 240 could communicate with satellite signal receiver 242 which receives data signals from satellite 244 which in turn is in remote communication with satellite signal transmitter 246. Terminals, for example further processing system 248, notebook computer 250 or satellite telephone 252, can thereby communicate with network 202. A local network 260, which for example may be a private network, LAN, etc., may also be connected to network 202. For example, network 202 could be connected with ethernet 262 which connects terminals 264, server 266 which controls the transfer of data to and/or from database 268, and printer 270. Various other types of networks could be utilised.

[033] The processing system 100 is adapted to communicate with other terminals, for example further processing systems 206, 208, by sending and receiving data, 118, 120, to and from the network 202, thereby facilitating possible communication with other components of the networked communications system 200.

[034] Thus, for example, the networks 202, 230, 240 may form part of, or be connected to, the Internet, in which case, the terminals 206, 212, 218, for example, may be web servers, Internet terminals or the like. The networks 202, 230, 240, 260 may be or form part of other communication networks, such as LAN, WAN, ethernet, token ring, FDDI ring, star, etc., networks, or mobile telephone networks, such as GSM, CDMA, 4G, 5G etc., networks, and may be wholly or partially wired, including for example optical fibre, or wireless networks, depending on a particular implementation.

Growth estimator

[035] Prediction of longterm growth of an asset can be difficult. Such a prediction maybe an estimate of change to the price of an asset, or selection of one asset over another based on performance of each asset over a period of time. An alternative is to predict the performance of the asset when compared to a collection of other assets. One example of such an asset is real estate properties, such as an individual property compared to other properties located nearby or in surrounding suburbs. However, selection of an asset may not only be related to the rank or long term relative growth potential of the asset. The asset may also need to meet additional criteria for selection. An example of such a criteria is when a growth estimator selects an asset when a predicted rank exceeds a threshold rank. The threshold may be determined using historical, or back testing, data to determine a threshold to meet an investment criteria of a commercial entity.

[036] Prediction of the long term relative growth of an asset and application of an investment criteria may be achieved using a growth estimator, such as a gradient boosted machine model, applied to features back tested to show capability at predicting long term relative growth. A growth estimator, such as a gradient boosted machine, learns which features, and combination of features, can be used as predictors of long term relative growth. While the expression "long term" is used in the specification, long term may be different for different asset classes. Long term, for the purpose of the growth estimator, is set by a parameter that is input to the growth estimator. If the asset is a domestic or commercial property a long term investment term may be 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, or more years.

[037] A growth estimator generation process 300 will now be described in relation to Figure 3. The growth estimator generation process 300 trains a growth estimator for use in ranking and/or selecting an asset. Typically the growth estimator is trained using historical information for the asset type or asset class. Data relating to the asset may then be applied to the trained growth estimator and a long term relative growth estimate made for the asset for the period of the investment term. Typically a growth estimator is generated for a set investment term. If the investment term is modified, then a new growth estimator may be trained for the modified investment term. The growth estimator generation process 300 may be practiced on a computer such as the processing system 100 communicating over a network 202 or on dedicated hardware.

[038] The growth estimator generation process 300 starts with a collect training data step 310 where data is collected for training of the growth estimator. The training data should have some relevance to the asset type and have historical values that can be used for training and assessment purposes. The training data may be a historical features related to the asset type and include information related to the asset type. Examples of training data used as historical features will be described below. Types of data that may be used for training, as well as operation of the growth estimator, include asset data relevant to the asset type, market data which is relevant to the state of the market for the asset type as well as high level data such as macroeconomic data.

[039] At a training step 320 the training data collected at the collect training data step 310 is organised as features and used to train the growth estimator. That is, a machine learning model is generate for the asset type using the historical features. In one example, the growth estimator is a gradient boosted machine and the trained model is a gradient boosted machine model. Typically only a portion of the training data collected from the collect training data step 310 is used in the training stage, with the remaining data used for validation of the trained growth estimator. Typically 60-90% of the data collected is used in training with the percentage of data varying based on the amount of training data available.

[040] At a determine threshold step 330 a threshold is determined for use by the growth estimator to select suitable assets. An investment criteria may be translated into a threshold by using historical, or back testing data, to determine a reasonable threshold that balances (i) an amount invested by a co-investor, relative to the total amount borrowed for the asset, (ii) a return required and (iii) expected underperformance, being a probability of underperformance multiplied by a magnitude of underperformance given an underperformance event occurs. Options for the threshold may be presented to a user of the growth estimator generation process 300, along with the three aforementioned summary statistics. In one example, the user may be presented with a value for an amount invested relative to the total amount borrowed, the return required and the expected underperformance. The user may vary one or more of the three values and see the threshold being recalculated to allow for exploration of the interaction between the three variables, as well as an effect on the threshold.

[041] At a deploy system step 340 the trained growth estimator is deployed for use on an asset. The trained growth estimator estimates the long term relative growth by applying features related to the asset to the growth estimator. The asset may be selected as a suitable investment based on the threshold from the determine threshold step 330. If the growth estimator determines a long term relative growth that exceeds the threshold then the asset is determined as being suitable. Once the long term relative growth of the asset is determined a report may be made with relevant information on the asset as well as how the long term relative growth did in relation to the threshold. The relevant information may be information on the asset. If the asset is a property then the relevant information may include property location, property type, information about the property and historical sale information for the property.

[042] An example of features that may be used as training data will be described in relation to Figure 4 which shows a feature table 400. The features of the feature table 400 may be the features collected at the collect training data step 310 of the growth estimator generation process 300.

[043] The feature table 400 has an asset identifier 410, which may be used as a primary key in a database. The asset identifier 410 may be unique for each sale of the asset. For example, in the case where the asset is property, a sale of a property in the year 2010 will have a different asset identifier 410 to a sale of the same property in the year 2005.

[044] A years held 420 field records how long the assets was held, which is equivalent to the investment term. In the feature table 400 all assets shown were held for 6 years. For a feature vector used as input to the decision for an asset under consideration, the years held field may be a value of how long the asset may be held expressed as a value such as years, months or days. A growth percentile 430 field forms part of the feature table 400. The growth percentile 430 is a value showing the relative growth of the asset, with the asset shown in the first row having an average growth percentile of 99, while the asset of the second row has an average growth percentile of 87. If the features of the feature table 400 are used to determine a long term relative growth of an asset, then the growth percentile 430 is unknown.

[045] Determining the growth percentile may be done from the sale price information. For years where there is no sale price information, an asset sale price may be created by interpolating, or extrapolating the sale price using the median price for the asset. In the example where the asset is a property, the median price may be for sales in the suburb. When there is sale price before and after the current year then the sale price may be interpolate between the historical transactions using the movement of the median price or extrapolated from a most recent transaction. The years held 420 may be used to determine a growth percentile 430 for a number of years, based on asset price growth calculated using the interpolated or extrapolated prices. Such an approach allows a value of growth percentile 430 to be calculated for all years within a time period. For example the growth percentile 430 may be calculated for an asset for 2020, 2019, 2018, 2017 and each year to 2010. The growth percentile 430 for each year would be for where the years held 420 value is 6 years. The average annual growth rate may then be converted to the growth percentile 430.

[046] Additional features may also be included, such as additional feature 1440, additional feature 2 450, additional feature 3 460 and additional features 470. Examples of additional features when the asset is a property may include, but are not limited to, median asset value in a suburb, median asset value in street, median asset value in a segment, where segment refers to comparable assets that may be defined by suburb, asset type, number of bedrooms, number of bathrooms and number of car spaces. A ratio of median asset value in the street to the median asset value in the suburb, a ratio of the median asset value in a segment to the median asset value in the suburb, a ratio of asset value to median asset value in the suburb, a ratio of the asset value to the median asset value in the street, a ratio of the asset value to the median asset value in the segment, a change in any of the previous data types from last year to the current year, a change in any of the previous data types from before last year to last year, time since renovation to the asset where the renovation exceed a threshold value of the property such as 5%, 10%, 15%, 20%, 25%, 30%, 35%, 40%, 45% or 50%. Another data type is months since construction of the asset.

[047] An asset evaluation process 500 will now be described with reference to Figure 5 where a growth estimator model is trained for an asset type and used to evaluate an asset.

The asset evaluation process 500 is applied to evaluate long term relative growth of the asset being evaluated.

[048] A number of inputs are used by the asset evaluation process 500 and includes asset information input 510, time horizon input 515 and data input 540. The asset information input 510 is information relating to the asset. In one example, when the asset is a property, the asset information may be the property address. The time horizon input 515 is investment term/how long the asset is predicted to be held and may define the long term aspect related to the long term relative growth of the asset. The data input 540 is additional information related to the asset such as the additional information set out above in relation to the additional feature 1 440, additional feature 2 450, additional feature 3 460 and additional features 470 of the feature table 400 described above. The data input 540 may be data from first and/or third part data sources.

[049] The time horizon input 515 may be used to filter the data input 540 based on the investment term when training the growth estimator. The time horizon input 515 means that the growth estimator is predicting how the asset will perform over the investment term, such as six years. The data input 540 is manipulated to select relevant data and to produce features, such as the features described in relation to feature table 400. For training purposes, with a time horizon input 515 of six years, the feature values from six years ago are compared against a response value for today, the feature values seven years ago are compared against a response value last year, the feature values from eight years are compared against a response value two years ago, and so on. Feature values for some years may interpolated or extrapolated from existing data, as described above. The historical portion of the data input 540 is sent to trained model input 550 for training purposes, as described above in relation to the training step 320.

[050] The asset information input 510 and data input 540 are combined at the model scoring input 520 to form features used by the trained growth estimator model to determine long term relative growth of the asset over the investment term set by the time horizon input 515. Typically the input information is combined into a vector, similar to the information described in relation to the feature table 400 of Figure 4, without the growth percentile 430 information. The information may be combined from multiple data sources using address information to selected data from different sources. The asset information input 510 is combined with the data input 540 to generate the features to create the model scoring input

520. The model scoring input 520 may use present day information and is applied to the trained growth estimator to determine long term relative growth of the asset.

[051] A trained model input 550 is used as input to the asset evaluation process 500 after being trained with historical data from the data input 540. The trained model input 550 may be the model produced by the growth estimator generation process 300 and is configured to process the model scoring input 520 at an apply model step 530. The data collected by the model scoring input 520, including the asset information input 510 and the data input 540, are processed by the trained model input 550 at the apply model step 530 to determine a long term relative growth percentile for the asset. The long term relative growth may be a predicted percentile value of growth for the asset and is for the asset after a fixed time period, the time period being the investment term.

[052] At threshold test 560 the long term relative growth percentile from the apply model step 530 is compared to a selection threshold input 517. The selection threshold input 517 is determined earlier at the determine threshold step 330 of the growth estimator generation process 300. If the output of the apply model step 530 exceeds the selection threshold input 517 then the asset evaluation process 500 progresses to asset accepted 570 and investment in the asset may proceed. If the out of the apply model step 530 does not exceed to the selection threshold input 517 the asset evaluation process 500 progresses to asset rejected 580 step where investment in the asset may be declined. Information relating to the output of the apply model step 530 and the threshold test 560 may be presented to a user of the asset evaluation process 500.

[053] An asset evaluation process 600 will now be described in relation to Figure 6. The asset evaluation process 600 may be practiced on a computer such as the processing system 100 communicating over a network 202 and takes information relating to the asset to determine if asset is a suitable investment. The asset evaluation process 600 is similar to the asset evaluation process 500 described above, however the asset evaluation process 600 does not involve any training of the growth estimator. The asset evaluation process 600 and the asset evaluation process 500 may share common processes.

[054] The asset evaluation process 600 starts with a receive asset information step 610 where information about the asset to be assessed is received. The information may be an address if the asset type is a property. At a receive data step 620 data related to the asset is received. The data from the receive data step 620 may be processed at a generate features step 630 where the data may be filtered or modified to produce suitable features, such as the features described above in relation to some of the columns of the feature table 400, such as the additional feature 1 440, the additional feature 2 450, the additional feature 3 460 and the additional features 470.

[055] At an apply features to growth estimator step 640 the features generated at the generate features step 630 are input to the trained growth estimator. The output of the growth estimator may be a long term relative growth measure over the investment term for which the growth estimator was trained. As described above, the growth estimator may be trained to predict growth over an investment terms such as ten years. The threshold, as determined at the determine threshold step 330 of the growth estimator generation process 300, is applied at the apply threshold step 650 to determine if the long term relative growth estimated by the growth estimator is suitable for investment.

[056] The asset evaluation process 600 concludes by outputting the result to a user of the system. The results may be displayed as field on a user interface where the user enters the asset information and the asset evaluation process 600 displays if the asset is accepted or rejected for investment. Alternatively, the display may show the long term relative growth as a percentile, in addition to displaying if the asset is accepted or rejected. Other information may also be displayed. In one alternative, a list of assets may be submitted and a table of results displayed to a user of the asset evaluation process 600.

[057] A training system 705 and an asset assessor 750 will now be described. Each of the training system 705 and the asset assessor 750 may be practiced on a computer such as the processing system 100 communicating over the network 202. The training system 705 and the asset assessor 750 may execute some or all of the steps and processes described above in relation to the growth estimator generation process 300, the asset evaluation process 500 and the asset evaluation process 600.

[058] The training system 705 will now be described with reference to Figure 7A. The training system 705 has a data receiver 710 for receiving data over the network 202. The data is then used by a feature generator 715 to generate features for training the growth estimator as described above. The features are then used by the growth estimator trainer 720 to train the growth estimator. Finally a threshold generator 725 may set as described above.

[059] The asset assessor 750 will be described with reference to Figure 7B. The asset assessor 750 receives information relating to the asset at a data receiver 755. The information may be processed by a feature generator 760 to generate features. The features are used as input to a growth estimator 765 where the growth estimator trained by the training system 705 is executed. The asset assessor 750 also includes a threshold tester 770 where the threshold determined by the threshold generator 725 is applied to the output of the growth estimator 765. The threshold tester 770 allows the asset assessor 750 to determine if the asset is suitable for an investment.

Variations

[060] While the growth estimator has been described as a gradient boosted machine, other boosting methods may be used or other machine learning algorithms. Examples of alternatives boosting methods include AdaBoost algorithm and the LightGBM algorithm. Example of alternative machine learning algorithms include neural networks.

[061] The growth estimator may be implemented using an extreme gradient boosted machine such as xgboost, an optimized distributed gradient boosting library. The growth estimator trained at the training step 320 may be trained with a linear or tree model as the booster and the evaluation metric as mean absolute error. The model may be manually hyper tuned to determine the most appropriate parameters for configuring the xgboost model. Parameters such as minchild-weight, max_depth, eta, subsample and earlystoppingrounds.

[062] In one example, the growth estimator may be a classifier which classifies an asset as an investment or not an investment. In such an example the threshold is not determined separately to the training of the growth estimator. The operation of the threshold may be incorporated into the growth estimator.

[063] The model developed for the growth estimator may be applied to all assets within an asset class. When the asset is a property, the same model may be used for all properties located within a geographical area, such as a country, state, local council area, suburb or street. For alternative asset types, a single model may be used for all assets of the asset type.

[064] The growth estimator described above is trained based on an investment term, such as the time horizon input 515 of the asset evaluation process 500. A change in the investment term may require the growth estimator to be retrained with the new investment term. In one example of the growth estimator, the investment term is not used as an input when assessing an asset as the investment term is incorporated into the growth estimator during training. In an alternative example, the investment term is an input in to the growth estimator when assessing an asset.

Advantages and Interpretation

[065] The described asset selection engine may provide an advantage that the growth estimator, such as a gradient boosted machine, learns the key features to predict relative long term growth and allows for differentiation by location and asset type. Different locations, asset types or a combination of the two may use different features that can change over time. The use of a growth estimator trained on existing data can take into account such differences.

[066] One advantage of the asset selection engine is the use of relative growth and the use of an output based on relative growth. Relative growth may require far less accurate macroeconomic predictions than predicting absolute growth. As such, relative growth predictions may provide for greater accuracy compared to an absolute growth prediction as accurate macroeconomic predictions can be difficult.

[067] One advantage of the asset selection engine is that predicting relative long term growth may assist in selection of assets within a top X% of a market. Such an approach may allow investment scale during both up and down markets.

[068] The reference in this specification to any prior publication (or information derived from the prior publication), or to any matter which is known, is not, and should not be taken as an acknowledgment or admission or any form of suggestion that the prior publication (or information derived from the prior publication) or known matter forms part of the common general knowledge in the field of endeavour to which this specification relates.

[069] Throughout this specification and the claims which follow, unless the context requires otherwise, the word "comprise", and variations such as "comprises" or "comprising", will be understood to imply the inclusion of a stated integer or step or group of integers or steps but not the exclusion of any other integer or step or group of integers or steps.

Claims (10)

The claims defining the invention are as follows:

1. A method of selecting an asset for investment, the method comprising: receiving a plurality of historical features related to an asset type of the asset, the historical features including information related to the asset type; generating a machine learning model for the asset type using the plurality of historical features; estimating a long term relative growth for an asset by applying the features related to the asset to the machine learning model; and comparing the estimated long term relative growth to a predetermined threshold to select the asset as an investment.

2. The method according to claim 1, wherein the long term relative growth is a predicted percentile value of growth for the asset.

3. The method according to claim 2, wherein the predicted percentile value for the asset is for the asset after a fixed time period.

4. The method according to any one of claims 1 to 3, wherein the predetermined threshold is determined using historical data.

5. The method according to claim 1, wherein the plurality of historical features related to the asset type comprise at least one data type selected from a list of data types consisting of asset data, market data, and high level data.

6. The method according to claim 1, wherein the predetermined machine learning model is a gradient boosted machine.

7. The method according to claim 6, where in the gradient boosted machine is an extreme gradient boosted machine.

8. The method according to claim 1, wherein the machine learning model is generated based on a fixed time period for the investment term.

9. An apparatus for selecting an asset comprising a processor, wherein the apparatus is configured to: receive a plurality of historical features related to an asset type of the asset, the historical features including information related to the asset type; generate a machine learning model for the asset type using the plurality of historical features; estimate a long term relative growth for an asset by applying the features related to the asset to the machine learning model; and compare the estimated long term relative growth to a predetermined threshold to select the asset as an investment.

10. A method for generating a machine learning model for estimating long term relative growth of an asset type, the method comprising: receiving a plurality of historical features related to the asset type, the historical features including information related to the asset type and being selected based on a fixed time period for an investment term; and generating a machine learning model for the asset type of the asset and the fixed time period.